Conferences, Videos & Testimony
Report of May 21, 1999 Workshop
by Kevin L. O'Neill
May 21, 1999
report author: Kevin L. O’Neill with review and concurrence of David Albright, Chair of the Source Term Subcommittee Final draft submitted: June 21, 1999
On May 21, 1999 the Health Advisory Panel (HAP) Source Term Subcommittee, chaired by David Albright, convened a workshop to present final draft findings of plutonium releases from the September 11, 1957 fire in Building 71 at the Rocky Flats Plant. Presenters at the workshop included Michael Diliberto of Diliberto and Associates and Paul Voillequé of MJP Risk Assessment and the Radiological Assessments Corporation (RAC). The workshop was attended by approximately 20 members of the public, including several former plant workers. Some of these workers who had detailed knowledge of the plant, including the design, construction and normal operating practices of Building 71 at the time of the fire.
This report summarizes the presentations by Michael Diliberto and Paul Voillequé. Diliberto, a well known fire expert, presented many of the findings of his report, entitled “Rocky Flats Building No. 71 Plutonium Fire Analysis,” dated April 30, 1999, which had been provided to workshop participants earlier. He had been asked by the HAP to prepare an analysis of the spread of the September 1957 fire in Building 71. The analysis was necessary because uncertainties associated with the fire’s progression affect RAC’s estimate of the amount of plutonium released to the environment. Subsequently, Voillequé discussed the findings of RAC’s Final Draft Report, entitled “Estimated Airborne Releases of Plutonium During the 1957 Fire in Building 71: Verification and Analysis of Source Terms,” dated March 1999. The RAC report was also released to workshop participants.
Diliberto’s presentation and report focussed on the cause and location of an explosion that took place in the main filter plenum during the fire. In doing so, he also needed to describe the progression of the fire from its point of origin in a glovebox line in room 180 to the main filter plenum (consuming the booster filtration system, which filtered air from the glovebox line, in the process) and estimated the rate at which filters in the booster system and the main filter plenum were penetrated by the fire. Each of these issues affect the estimate of plutonium released as a result of the fire. Diliberto also touched upon operating practices at Rocky Flats that contributed to the fire’s development, spread, and severity.
Diliberto opened his presentation by briefly describing the layout of building 71 and its ventilation system. He provided a detailed schematic (“Exhibit 1” in the presentation) of the areas in Building 71 that were affected by the fire. In response to a question, Diliberto noted that he was unable to locate operating records of the building’s ventilation system, particularly records of how the fans were operated. Diliberto said that he relied on available design documents and pre-fire reports on how the fans and ventilation system were operated.
Factors Affecting the Fire’s Development
Diliberto then discussed the factors that contributed to the fire’s development. First, he noted that the plutonium in the glovebox line in room 180 was stored improperly. The glovebox line was not designed to be airtight, which allowed room air to flow into the glovebox line, increasing the exposure of the plutonium to oxygen and moisture and aiding in its spontaneous combustion.
Diliberto noted further that the plutonium in room 180 was “alpha phase” plutonium, which oxidizes more rapidly than the more commonly used “delta phase” plutonium. He said that there was general unfamiliarity on the part of plant workers with alpha phase plutonium, and that the plutonium involved in the fire had not been properly removed and stored in an inert container during non-operational times.
Workshop participants discussed this finding. According to one former plant worker who attended the workshop, delta phase plutonium forms a protective coating when it oxidizes, which prevents spontaneous combustion, while alpha phase plutonium does not. The worker also observed that at the time of the fire, there was little experience at machining plutonium. Machining chips, filings and residues are more susceptible to combustion because they have relatively larger surface areas that are exposed to oxygen. The worker said that there were “lots of small, contained fires” in the glovebox line prior to the September 1957 fire, including a fire at the lathe one month earlier.
Diliberto then noted that plant workers had a lack of understanding about the fire risks associated with plexiglas and other materials associated with the construction of the gloveboxes. He noted further that plexiglas was widely used to construct glovebox enclosures, partitions and view boxes. He said that the lack of inert firebreaks between sections of the glovebox line contributed to the origin and spread of the fire.
A member of the public asked about the flammability of plexiglas used to construct glove boxes. Diliberto said that he was not able to determine the grade of plexiglas used at the time of the fire. However, he said that a 1970 test of different combustible materials found that many grade of plexiglas are highly flammable. A plant worker said that there was no standard fire-resistant plexiglas used prior to the fire.
Another member of the public asked why leaded glass was not used to make gloveboxes. In response, a plant worker said that leaded glass was used in plutonium production facilities, but not for research and development activities such as those taking place in room 180.
Factors Affecting the Fire’s Spread and Severity
Next, Diliberto discussed several factors that contributed to the fire spread and its severity. Among the factors cited by Diliberto were the lack of compliance by the plant with recommendations published in either AEC papers or National Fire Protection Association (NFPA) guidelines that were in effect at the time. In particular, he noted that NFPA standards on exhaust system maintenance relating to the frequency of inspection and cleaning of ducts and filters were not followed at the plant.
As a result of failing to follow maintenance practices, Diliberto concluded that the chemical warfare system (CWS) filters used in the booster system and main filter plenum became heavily loaded with combustible dust (6 to 10 lbs., on average). These filters—composed of 86 percent cellulose and 14 percent asbestos—were themselves combustible, as were the filter frame, rubberized gasket and adhesive bond between the filter medium and the frame. Diliberto provided a figure to illustrate the construction of one of the CWS filters used in the booster system and the main filter plenum.
Diliberto noted that one third of the 620 filters in the plenum were destroyed as a result of the fire. A participant asked about the certainty of that estimate, noting that “many filters” were pulled out of the fame by fire fighters to form a firebreak. A plant worker concurred with Diliberto’s estimate, noting that only a few filters were pulled out of the frame. He stated that the filters were difficult to remove because they were bolted to the plenum frame.
Diliberto then discussed numerous controlled tests that were conducted to determine the rate of combustion of CWS filters similar to those in use at Rocky Flats at the time of the fire. These tests were conducted at Hanford, Rocky Flats and Lawrence Livermore Laboratory in the 1950s and 1970s. Numerous tests, particularly those conducted at Rocky Flats in 1957 and 1959-1960, were conducted on CWS filters that were saturated with dust or with chemicals. Later tests, held at Lawrence Livermore, were conducted on HEPA filters to demonstrate “smoke plugging.” Diliberto noted that these filters had similar characteristics as CWS filters, except for combustibility. Diliberto presented a detailed, annotated figure plotting the temperature curve over time for the Hanford and Rocky Flats fire tests (“Exhibit II”).
Diliberto said that the key relevance of these tests for the 1957 fire investigation was to help determine both the rate of filter penetration by the fire and to determine the cause of the explosion. One test series in particular, conducted at Rocky Flats in 1957, showed a filter burn through in approximately 2.5 minutes. However, Diliberto said that there was no documentary evidence (films, photographs) of that particular test.
Significantly, Diliberto noted that in many of the tests, small “flash back” explosions emerged from the face of the filter. Diliberto explained that these explosions were caused by the release of pockets of combustible gases, such as carbon monoxide, that had become trapped in the burning filter. These gas pockets were trapped as the filter became plugged with debris, soot and smoke particles. When the flame reached these pockets, a small flash back resulted.
Diliberto then discussed a fire spread and development time line. First, he introduced the timeline of events given in the fire report prepared by the plant after the fire. Several of the events Diliberto described elicited comments and clarifications from participants:
- One plant worker offered a point of clarification to Diliberto’s presentation. Diliberto understood the boiler operator’s observation at 10:28 PM that there was “smoke coming from the building’s main exhaust system stack” to mean that the operator saw smoke leaking through the clean side of the filter plenum. However, a plant worker noted that the worker had access to a platform outside the building and could have seen smoke leaving the stack itself.
- One participant observed that water was introduced in room 180 to extinguish the fire at 10:38 PM. This participant asked if a metal water reaction could have generated sufficient hydrogen to cause the explosion in the plenum. Diliberto said that it was unlikely; when water was introduced in room 180, it was likely that the plutonium in the room had already burned. The water, he said, was used to extinguish burning combustibles (eg., plexiglas). A plant worker agreed, stating that there was not enough metal left in the room to produce sufficient hydrogen. This comment led to a discussion about the differences between how carbon dioxide and water extinguish a fire. Diliberto noted that carbon dioxide, which was first used by fire fighters to extinguish the fire, is a suffocating agent that is most effective on contained fires that are highly localized. However, as in the case of the fire in room 180, which he described as “class A combustibles burning freely in a large room,” carbon dioxide has little effect. In contrast, water acts as a cooling agent that extinguishes a fire by transferring the heat of the burning material from the burning object to the water, resulting in steam.
- During the presentation, Diliberto noted that the filter plenum was not constructed to withstand an explosion, only to handle “reasonable overpressures” generated by the exhaust fans operating at high speed. One participant stated that the filter bank should have been more sturdy. Diliberto agreed that he would have supported a sturdier construction, but noted that the lack of fire dampers in the ventilation system effectively reduced the effect of the explosion in the plenum area. Moreover, Diliberto said that if the plenum had been constructed with firebreaks, it would have entailed a different design on the exhaust fan system. A plant worker agreed with Diliberto, noting that the design criteria has changed since the 1950s. At the time, the worker said, the principal concern was preventing a criticality over all other concerns in the plant’s fire prevention / suppression plan.
To conclude this section of the presentation, Diliberto summarized the timeline in a detailed figure (“Exhibit IV”) that described the temperature - time curve for all combustibles involved in the fire. The key outcome of the sequence of events, Diliberto said, is that by the time of the explosion, the main filter plenum the entire ventilation system in the building was backed up with smoke.
Moreover, he noted that the filters in the main filter plenum were burning and were plugged by particulates, soot and smoke from the burning combustibles in room 180, the completely destroyed booster system, and by the burning filters themselves. He noted that filters in the main plenum began to heat up and may have started plugging by 10:18 PM. By 10:20 PM the filters would have ignited, and by 10:24 PM begun to rupture. At 10:25 PM the exhaust fans were put on high speed, and three minutes later smoke was observed coming out of the stack. The filters continued to burn and plug up, giving off small flash back explosions into the upstream side of the plenum. The shock of these small explosions released combustible dust into the plenum area, which was also consumed and contributed to the fire.
Participants discussed Diliberto’s conclusion about smoke-plugged filters. One participant observed that since only a small number of filters were being plugged, the pressure drop across the filter plenum remained high. Diliberto defended his conclusion by observing that placing the fans on high speed accelerated the filter plugging and rupture. He also noted that one of the fire tests at Lawrence Livermore, discussed previously, showed that plugging was greatest with plexiglas and other combustibles were introduced. A plant worker concurred, noting that there would be a “blowtorch” effect, given the localized nature of hot gases and embers entering at one end of the plenum.
Diliberto then illustrated the growth of the fire in the plenum with a series of figures prepared by Voillequé. He observed that at the start of the plenum fire at 10:18 PM, only 6 filters were initially penetrated. By 10:25 PM, when the fans were placed on high speed, 9 filters were penetrated. Within five minutes after the fans had been placed on high speed, the number of penetrated filters grew to 32. At 10:40 PM (at the time the exhaust fans went off and one minute after the explosion), the number of penetrated filters had frown to over 60, with the fire burning “horizontally and vertically across the face of the plenum.” Six minutes after the explosion, the rate of the fire expansion had slowed, with an estimated 68 filters penetrated.
Factors Contributing to the Explosion
Diliberto then identified and discussed the factors that contributed to the explosion in the plenum. These factors included:
- the lack of a automatic sprinkler system on the upstream side of the filter plenum;
- the lack of a fire detection system in the booster system and the deactivation of the system in the main filter plenum;
- the lack of fire dampers in the exhaust ducts where they terminate at the main plenum wall;
- the lack of maintenance and housekeeping of the main filter plenum, resulting in excessive accumulation of combustible dust and lint on the filters;
- the continued use of CWS-6 filters, even though more fire resistant CWS filters had been developed; and
- the “flashback” fire phenomenon, discussed earlier.
Diliberto’s list of factors elicited many comments from workshop participants. In particular, former plant workers argued that the tone of the presentation was accusatory; there were “good reasons for each of the ‘lack of’s.’” For example, at the time of the fire, circumstances and practices did not permit the installation of fire protection equipment or maintenance practices that later became standard. For example, a water sprinkler system was never installed because the Congress never appropriated funds to install such equipment. One worker also repeated an earlier comment that the primary factor governing fire protection was the avoidance of a criticality. As a result, the use of water was discouraged.
Another worker argued that filters in the plenum were not changed because they remained effective, even after their design operating periods. The worker said that individual filters in the plenum were regularly DOP tested to check their reliability, and Voillequé noted that the building’s monthly health physics reports made reference to the replacement of some filters. According to the plant worker, “the cold side of that plenum remained cold.”
Nonetheless, Diliberto’s list also elicited comments that production was the plant’s and the government’s priority. Safety was important, but often short-changed when it conflicted with the production priority.
Explosion Development Theories
Diliberto then presented alternate theories of how the explosion developed. The first theory asserts that the explosion occurred when explosive gases built up in the duct work and were ignited by the fire from within room 180. Diliberto explained that this scenario was unlikely, because by the time the explosion occurred, there was no source of ignition in either room 180 or the ducts.
A workshop participant asked if a spark from a damaged power supply line could have ignited combustible gas in the ducts. Diliberto acknowledged that the power lines may have been damaged by the fire and could have been sparking, but that there was no evidence of heavy damage of the ducts between room 180 and the plenum. If the explosion had occurred in that area, then these ducts would have been heavily damaged.
Diliberto offered a second theory, which he also quickly discounted. Under this theory, a metal-water reaction produced hydrogen, which exploded in room 180. The hydrogen was generated when water was used to extinguish the fire in room 180. Diliberto said that this theory was also unlikely; personnel in room 180 would have been “well aware” of an explosion that had occurred in the room itself. The theory also was not considered in the December 1957 supplementary fire report. Diliberto also reminded participants that the explosion occurred after the fire in room 180 was extinguished and, moreover, the bulk of the burning materials in room 180 was not plutonium metal.
Primary explosion theory
Diliberto then presented the “primary explosion theory”—the flashback phenomenon in plenum filters ignited heavy concentrations of carbon monoxide (CO), lint and dust in the air on the upstream side of the plenum filter bank. The CO buildup resulted primarily from incompletely burned filters in the plenum. Filter plugging prevented the removal of CO from the main filter plenum, even as flash back explosions vibrated the plenum, which suspended built up dust and lint from the surface of the filters.
To support his conclusion, Diliberto observed that the fuel supply (CO, dust and lint) was in place—it did not have to be transported to the plenum areas. He also noted that plutonium was blown back into the building through the ducts as a result of the explosion. He also asserted that calculations indicated that a pressure pulse sufficient to generate an explosion of the magnitude experienced in room 180 would not cause “gross damage” to the plenum itself.
These conclusions generated much discussion among participants. One former plant worker, supported by several participants, argued that sufficient CO could not have built up in the clean side of the plenum because the CO would leave the plenum through non-plugged filters. Diliberto asserted, however, that calculations showed that the rate of CO buildup, generated by the burning filter plenum, would have produced CO exponentially over time, while the amount of CO removed from the plenum would have occurred linearly. He also noted that an explosion that produces a pressure pulse on the order of 1 pound per square inch (PSI) was “not a very large explosion.” He presented a table showing that conditions were sufficient in the plenum to produce sufficient CO within the plenum during the fire sequence presented earlier to produce a 1 PSI explosion.
Diliberto asserted that the explosion originated in the west end of the plenum, where the duct outlets from the booster system and from room 180 were located. This conclusion was also scrutinized by workshop participants. One former plant worker observed that the greatest damage to the ducts was on the downstream side of the filter plenum in the area leading to the exhaust fans, the exhaust fan housing, and to one of the fans, itself. In contrast, the ducts upstream to the main filter plenum were relatively undamaged. Given this evidence, the worker asked if the explosion could have taken place in the downstream side of the filter bank or in the fan housing itself.
Diliberto said that this scenario was unlikely since the operating exhaust fans would have prevented CO buildup in that area. Even if the fans were not operating, the supply fans would still be pushing CO through the fan outlets. Voillequé and some workshop participants agreed with Diliberto’s conclusion.
A HAP member asked if it was possible that two explosions had taken place. Two different events could lead to the asymmetric damage to the ducts. Diliberto said that the fire reports referred to only one event; if there were two events had occurred, they would have to have occurred simultaneously. He thought this was unlikely.
One member of the public observed that “explosions often give results that are difficult to explain.” However, he believed that a criticality in the plenum caused the explosion. Diliberto, Voillequé and some workshop participants said that there was no evidence (neutron irradiation, detection of fission products) to support a criticality theory. Voillequé, Diliberto and former plant workers also asserted that there was insufficient plutonium in the plenum to produce a criticality.
A HAP member asked if the damage to the fans could be explained by the high heat of the fire. Diliberto agreed, noting that the temperature of the air in the downstream side of the filter that was exiting through the fans would be much hotter than the temperature on the upstream side. It was also observed that the lead lining at the top of the stack was also deformed by heat. This answer seemed to resolve the question to the satisfaction of the plant worker who raised the issue and to many of the other workshop participants.
Following Diliberto’s presentation, workshop participants turned their attention to Paul Voillequé, who discussed the principal findings of the RAC report entitled “Plutonium Release Estimates for the 1957 fire in Building 71.” Because Diliberto’s presentation had elicited so much discussion, and because many workshop participants were already familiar with Voillequé‘s general approach to estimating plutonium releases from the fire, Voillequé‘s presentation was shorter than planned. However, workshop participants offered many insightful comments on his report.
Voillequé began his presentation by briefly discussing the release pathways of plutonium involved in the fire. In particular, he noted that the plutonium in the room 180 glovebox line oxidized and traveled through the booster system and through the main room ventilation system to the main plenum. Plutonium previously deposited in the booster system filters and in the main filter bank also were released when the fire spread to those areas. However, Voillequé remarked that plutonium deposition in the ventilation ducts themselves contributed little to the overall release.
Voillequé then discussed the amounts of plutonium in the glovebox line that were involved in the fire. He noted that the 1957 fire report states that the approximate initial inventory of plutonium in the room was approximately 42 kg. However, not all of this plutonium was located in areas that were most affected by the fire. According to the accident report, 1.2, kg of plutonium contained in oils and sludge, 0.4 kg of plutonium metal pieces, 0.8 kg of casting residues, and 2.6 kg of plutonium oxide were in areas that burned most severely. He also said that, according to the accident report, 8 plutonium hemispheres of unknown mass were also in this area.
Based on this accident report inventory, Voillequé then estimated the quantity of plutonium that could have been oxidized. He estimated that 0.6 to 1.8 kg of plutonium in organic liquids and sludge (presumably including casting residues) and 8 to 16 kg of plutonium metal (including the hemispheres) were oxidized.
A participant asked if the casting residues should be included in the organic or if this source term should be considered separately, since the fire ignited through the spontaneous combustion of these residues. A second participant noted that plutonium contained in volatile chemicals, such as carbon tetrachloride (CCl4), should also be considered separately, since CCl4 is more volatile than organic liquids. Voillequé said he would address these comments in drafting the final version of the report.
Voillequé also was asked if it was possible to obtain the declassification of the total mass of the hemispheres involved in the fire. According to a HAP member, not all of the hemispheres were the same dimensions, and therefore may have different masses. He then observed that the Energy Department might be willing to declassify the aggregate mass of these hemispheres.
Voillequé was also asked if all of the plutonium in the hemispheres went into the air (ie., “burned” in the fire). Voillequé said that “piles of oxide” were recovered by plant workers after the fire, implying that at least some of the plutonium in the hemispheres stayed in the room.
Voillequé then discussed how he estimated the quantities of plutonium that were located in the booster system and the main filter plenum filters at the time of the fire. He noted that the plant implemented a duct sampling program for many of the ducts in the building, principally ducts that exhausted areas where plutonium handling and processing was the greatest. This sampling program also included the duct between the booster system and the main filter bank. Samples were collected on a daily basis.
Voillequé said that this sampling program involved only a single point sampler. Consequently, the plant data was not representative of the true concentration of plutonium in the ducts and this underestimated the loading on the filters. However, he had developed a correction factor for the report on routine plutonium releases from the plant, which uses Monte Carlo calculations on later plant data collected from three samplers to determine relative plutonium concentrations throughout the duct.
Voillequé noted that the amount of plutonium that collected on the booster system could not be determined the same way, since the ducts entering the booster system were not sampled. However, given an estimate of the filter efficiency of the booster system, and using sampling data collected from the booster system’s exhaust vent, an estimate could be calculated.
To account for plutonium releases to the plenum from rooms that were exhausted by ducts that were not regularly sampled, Voillequé used data collected by room air samplers. Based on this data, he concluded that a very small amount (less than 1.4 g) of plutonium had been released to the plenum from these rooms.
Voillequé then summarized his findings of the quantity of plutonium that had been released to the main filter plenum prior to the fire. He observed that several accidents and operations in the building accounted for most of the estimated release to the filters. The largest release, involving an estimated 300 grams, occurred during a June 1957 peroxide explosion in room 146. However, Voillequé observed that the duct from room 146 entered the opposite end of the plenum from where the September fire occurred. Therefore, this plutonium likely contributed little to the overall release during the fire.
Having determined the quantities of plutonium involved in the fire in room 180, the booster system and main filter plenum, Voillequé then discussed how he estimated the quantity of plutonium that actually became airborne as a result of the fire. He described how airborne release fractions (ARFs) had been developed for fires involving plutonium in various forms and different media (organic liquids, cellulose and metal). In addition, for plutonium metal, ARFs had been determined experimentally for different fire conditions (dependent on temperature, surface area of the metal, and other conditions), as well. A plant worker observed that an ARF for plutonium in hydrocarbons might also need to be considered.
Voillequé observed that the conditions of the burning plutonium metal remained uncertain. In particular, he noted that it seemed unlikely that a large portion of the plutonium was subject to the most vigorous fire conditions, given the observation of plant workers after the fire. To account for this uncertainty, Voillequé estimated releases for scenarios where different quantities of plutonium metal were subject to different fire conditions. As a baseline, he assigned equal quantities of plutonium metal to four different fire conditions. In his presentation, Voillequé noted that the assignment of different quantities of plutonium metal subject to different fire conditions remains a key uncertainty in determining the quantity of plutonium released to the atmosphere.
Using the filter failure time estimates, the supplemental chronology and the fire development theory, plutonium release estimates complied over fifteen minute intervals have been developed. These intervals are needed by other RAC investigators who are estimating the dispersion of airborne plutonium released by the fire to the surrounding area. Meteorological data needed for this estimate was recorded on a 15-minute basis.
Voillequé concluded his presentation by reporting his findings of the amount of plutonium released by the fire. Assuming a uniform distribution of plutonium metal into the four categories of metal oxidation, Voillequé estimated that a total 290 grams of plutonium was released. He noted that the uncertainty range was quite broad, with 90 percent confidence interval (5th and 95th percentiles) of 160 to 490 grams.