Cleanroom fire-safety debate heats up
By John Haystead
For semiconductor manufacturers, there`s no dispute that the cost of downtime, repairing and/or replacing multibillion dollar manufacturing facilities damaged by fire and/or smoke can be catastrophic, but there`s far less agreement between the industry, its standards organizations and insurance carriers on how to best address the threat.
Last November, the Factory Mutual Engineering Corp (Norwood, MA) proposed that a new set of testing requirements, together with a list of “fire-safe” materials, be incorporated into the latest interpretation of the National Fire Protection Association`s (NFPA) NFPA-318 Standard for the Protection of Cleanrooms. Factory Mutual (FM) is the engineering arm of three major commercial and industrial-property insurance companies (Allendale Insurance, Arkwright and Protection Mutual Insurance), who own or direct its various organizations, including the Factory Mutual Research Corporation (FMRC) which developed the new “Cleanroom Materials Flammability Test Protocol (FMRC-4910).”
The 4910 protocol defines a set of testing requirements and procedures as well as minimum performance requirements for evaluating materials used in semiconductor cleanrooms. Parameters include a Fire Protection Index (FPI), Smoke Damage Index (SDI) and Corrosion Damage Index (CDI). Although the FM proposal was voted down by the NFPA membership, and was similarly not included in the body of the Semiconductor Equipment and Materials International (SEMI) organization`s SEMI-2697 fire-protection document (included only as a referenced guideline), debate continues over the merits, practical implementation and future of the FMRC-4910 protocol.
What`s at issue?
The greatest potential impact of FMRC-4910 is on semiconductor process tools, particularly wet benches. As it stands now, FM is recommending that wet benches either be made from 4910-tested, fire-safe materials or that the tools incorporate integrated fire detection and suppression systems in addition to whatever fire-suppression systems are installed in the facility.
Not everyone agrees that such extensive measures are warranted. Although acknowledging that several tool-related fab fires in the Far East resulted in major losses, Steve Burnett, SEMATECH manager of environmental safety and health best practices, points out that, “fire-safety conditions in the U.S. are very different, and to our knowledge there hasn`t been the history in recent years of fires and other losses that would warrant drastic, costly changes and modifications to our systems.” Dave Namyst, risk engineering manager for Intel Corp., agrees. “It`s not appropriate to evaluate fire-safety practices and systems in the U.S. together with those of all other countries in the world. Many countries` practices and standards may be 10 to 15 years behind U.S. requirements in many areas,” he says.
Nevertheless, the immediate concern for semiconductor manufacturers involves the thousands of wet benches already in use. As pointed out by Dave Quadrini, chairman of the NFPA-318 Committee and corporate fire marshall for Texas Instruments, since it`s not possible to convert all existing fab tools to fire-hardened materials, the installation of integrated fire-suppression becomes the only option.
The Semiconductor Industry Association (SIA) was to decide last month at its meeting in Washington D.C. whether to contract for an analysis of existing tools to determine guidelines and options for retrofitting fire-suppression systems. The SIA has been working together with SEMI, SEMATECH, NFPA and FM to address changes to codes and industry requirements for fire safety and prevention.
It`s the process
For next-generation wet benches, FMRC-4910`s proposed use of fire-safe materials also raises concerns, not the least of which is process contamination. Although Quadrini agrees that he and many others would prefer not to have additional fire-suppression systems at the tool level “because of the collateral damage they can cause and danger of inadvertent release,” he points out that the introduction of new materials into the process may be equally hazardous. Likewise, SIA Chairman Lee Neal says that while the SIA “is fully supportive of addressing fire-safety concerns, we will not support arbitrary recommendations or the mandated use of new fire-retardant materials. Any new material must not only be proven effective in reducing fire hazards, but must also be tested for its serviceability in the semiconductor process environment, addressing concerns such as contamination and resistance to chemicals.”
According to Roger Benson, a Factory Mutual senior loss prevention specialist, three materials have already successfully completed FMRC`s 4910-protocol testing including a fire-retardant polypropylene material from Compression Polymers and a rigid PVC material from Takiron of Japan, “with four other materials close to completion and several others in the test cue.” But, says Norm MacDonald, president of FlexHead Industries (Boxboro, MA), “in addition to costing up to five times as much as current materials, most of the fire-retardant materials proposed thus far are still experimental with no test data for particulate or outgassing contamination or serviceability.”
FM`s Benson acknowledges these concerns. “The industry has valid points and if I were a fab manager, I wouldn`t use any new materials either until I was sure they wouldn`t produce problems for my process.” He points out, however, that as geometries continue to shrink, manufacturers have already begun looking at building tools from different materials from purely process considerations. “Some of these materials are better in both areas,” he adds.
To address the issue, SEMATECH, together with the SIA and SEMI, have launched a project to study the process compatibility of fire-safe materials. Says Burnett, “this is an issue that we need to address whether we like it or not, and hopefully, we`ll have a better idea of our options by the end of the year.” The objective of the SEMATECH project is to develop a test methodology for evaluating process compatibility, but Burnett expects they will also test some individual materials along the way. “This is the other half of the equation,” says Burnett, “and assuming we can find materials that in addition to their fire-safety characteristics are also process compatible, cost-effective and offer proper structural integrity, we expect there will be very little controversy with regard to their use in new tools.”
Who`s test is it?
In addition to process compatibility issues, Quadrini says there were other reasons why NFPA voted against inclusion of the FMRC-4910 protocol in the latest NFPA-318 document. “There were still too many unanswered questions, particularly regarding the repeatability of the testing process and that it is proprietary to one testing laboratory.”
Quadrini is referring to the fact that FMRC is currently the only testing body set up to conduct 4910-protocol testing, a situation which also bothers Intel`s Namyst. “It`s not necessarily a bad standard,” says Namyst, “but it`s not in the best interest of the industry to incorporate 4910 into any kind of code as long as it`s sole-sourced, with one company testing materials to its own standard.”
FM`s Benson says they are now addressing this concern and are already in discussions with other testing laboratories regarding conducting FMRC-4910 testing and they are also working with the American Society for Testing & Materials (ASTM) (West Conshohocken, PA) for approval of the FMRC-4910 test process and apparatus. “It`s not going as fast as we would like, but we recognize that many people would be more comfortable if 4910 tests were also recognized and listed as an NFPA or UL fire test.” According to Benson, although some early estimates of the cost of the commercially available test apparatus were more than $200,000, more recent figures show the apparatus costing around $75,000 – $80,000. Still Benson says, “it`s not clear how far away we are from having another laboratory qualified to set up and run the test.”
Even if FMRC releases 4910 to other testing laboratories, however, there are still questions regarding its implementation. Says Namyst, “FMRC has not addressed the issue that even if they cross-license 4910 testing to other testing laboratories such as UL (Underwriters Laboratory), that this means those results would be acceptable to FM.”
According to Ken Rhodes, associate managing engineer at UL, “this is an age-old issue and it`s not likely that UL would get involved in a cross-certification agreement with FM. Although we frequently run tests for FM and vice versa, unless the protocol is adopted as an industry standard, FM would probably want to monitor any testing performed here.”
As acknowledged by Benson, there are indeed several possibilities with regard to the extent to which FM will accept the results of other laboratory testing, and in all cases, FM will still require material manufacturers to enter into an audit agreement with FM. According to Benson, although his “first choice would be that the results of people whom we have licensed to conduct the test would be accepted,” it`s also possible that FM might require FMRC to run a “quickie” test of some kind to verify test results or indeed that FMRC run its own complete tests.
Where there`s fire, there`s smoke
Many industry observers believe FM`s push toward the mandated use of fire-retardant materials is too heavily weighted toward the liability concerns of the insurance carriers as opposed to true fire-safety issues or the production realities of the semiconductor industry. For example, some have pointed out that the initial motivation for FMRC`s research into fire safety was to determine the largest risk/cost exposure areas to insurance carriers.
Since the results of this research indicated that downtime resulting from smoke contamination was the single largest contributor to carrier costs, this has been cited as the reason the insurance industry is now pushing for the use of materials that will produce little or no smoke from fire. “Although the FMRC had earlier advocated the use of fire suppression systems, they are now proposing a doomsday scenario in which such systems completely fail, together with arbitrary thresholds for allowable production of smoke from materials,” says MacDonald.
FM`s Benson disagrees with this assessment. “What we`re trying to do is develop a balance between process and protection and we recognize that it is the process that drives it.” Still Benson points out that there are business and risk realities that must be addressed. “Even with the best-engineered tool systems, things can still happen and everything we recommend to our clients is loss-driven.”
To illustrate their positions, Benson outlines a hypothetical scenario in which a tool-related fire would reach a certain size before it was extinguished by the facility`s sprinkler system, and in which case they would estimate the loss expectancy (LE) at $40 million. “In this scenario, we would judge that level of loss as unacceptable, and would recommend that $75,000 be spent on a fire-suppression system for the tool.” Ultimately, however, Benson says the business/risk decision remains with the insured. “Manufacturers can either accept our recommendations or assume the risk themselves by raising their deductibles.”
Who makes the call?
Since every semiconductor company is not insured by an FM company, although other carriers may follow its lead, its requirements and recommendations do not necessarily apply to all manufacturers. Still, regardless of carrier or liability issues, the extent to which FMRC-4910 protocols are adopted by semiconductor manufacturers may be more driven by state and local code enforcement officials.
For example, as pointed out by Intel`s Namyst, “we have our own internal requirements to meet codes, which are enforceable in whatever local jurisdictions we are in.” These codes are set in conformance with the Uniform Building Code (UBC) of the International Conference of Building Officials (ICBO) and the Uniform Fire Code (UFC) of the International Fire Code Institute (IFCI). Though UFC and UBC codes often refer to NFPA standards, not all sections of NFPA are necessarily adopted as code in all areas of the country, and as observed by Namyst, “it`s certainly possible that ICBO/IFCI may adopt 4910 as part of their requirements.”
For his part, Benson says he can`t predict what ICBO will do with regard to FMRC-4910, but notes that some fire officials among the UFC group were very adamant about having 4910 included, at least in concept, in the NFPA-318 document.
Ultimately, as observed by SEMATECH`s Burnett, each company will have to make these decisions for itself, “but hopefully we can reach a common solution with FM that represents the best practices in our industry and will be a practical approach that addresses the concerns of FM but that also makes sense for the semiconductor companies in terms of risk management.” According to Quadrini, the NFPA plans to run a “shorter than usual” revision cycle for NFPA-318 that will bring it back up for review by the year 2000.