Strategies and techniques for sample collection and analysis: experience from the Swedish PCB accidents.

A series of polychlorinated biphenyl (PCB) fires and explosion in PCB-filled capacitors and transformers is discussed. A sampling program followed by isomer specific determination of trace levels of polychlorinated dibenzofurans (PCDFs) and polychlorinated dibenzo-p-dioxins (PCDDs) is described. Data from a series of Swedish PCB accidents are given. In addition to PCDFs and PCDDs, we have also found a series of polychlorinated biphenylenes (PCBPs). Cleaning of contaminated areas was done by vacuum cleaning followed by high pressure washing.


Introduction
Polychlorinated biphenyls (PCBs) have been used extensively as a dielectric fluid in capacitors and transformers since the 1950s. However, the use pattern in different countries is different (Table 1). It was found that commercial PCB mixtures solidified around -20°C, consequently, in Scandinavia where most transformers are placed outdoors, mineral oil is the most commonly used dielectric fluid in transformers.
In 1978 it was reported by Buser, Bosshardt, and Rappe that polychlorinated dibenzofurans (PCDFs) were formed in 1 to 25% yield by the pyrolysis of PCBs in the presence of air (1). This paper also contains a warning for all accidental and planned burning of PCBs or PCB-contaminated wastes.
During the period 1981 to 1983 a series of accidents were reported in Scandinavia and the United States that can illustrate this risk. These accidents can be subdivided into three different categories: (a) PCB fires with mineral oil as the external energy source; (b) PCB fires with electrical energy as the external energy source; (c) explosions in PCB-filled capacitors.
Some of these accidents are listed in Table 2, and some of these fires are also discussed in this article. These accidents all resulted in the formation of PCDFs, and in some cases also polychlorinated dioxins (PCDDs), biphenylenes (PCBPs), and pyrenes (PCPYs). It is generally accepted that a risk evaluation should be based *Department of Organic Chemistry, University of Umea, S-901 87 Ume&, Sweden. on levels of the toxic 2,3,7,8-substituted congeners (Cl4-Cl6) (Figs. 1-3). Consequently, a highly sensitive and an isomer-specific analytical method is required. In Sweden an inofficial guideline for levels of PCDFs after PCB fires has been established: 2,3,7,8-tetra-CDF, 50 ng/m2; total level of tetra-CDFs, 100 ng/m; total level of C14-C18 PCDFs, 1000 ng/m2.

Experimental
Sampling, Extraction and Clean-up A wipe test sample was collected by using a filter paper or Kleenex tissue. The sampled area was normally 2 dM2. A dry wipe was followed by a wet wipe with water. An organic solvent was found to increase the background noise. The levels are given in amount per square meter (ng/m2).
The sample (dried filter paper or Kleenex tissue from wipe tests or soot sample) was spiked with 1-2 ng of 13C12-2,3, 7,8-tetra-CDD, '3C12-2,3,7,8-tetra-CDF and 13C12-octa-CDD. Thereafter the sample was treated with 10 mL of 1 M hydrochloric acid for 1 hr with shaking. The slurry was filtered by suction on a Buchner funnel, washed with water, and dried in the air.
The dried material was transferred to a Soxhlet extractor and extracted with toluene (about 100 mL) for 8 to 48 hr at a speed of 1 to 2 mL/min. The toluene was evaporated to 100 ,L in a Rotavapor with some external heating. The rest of the solvent was eliminated by a purified N2 flush without external heating. The residues were dissolved in 1 mL of n-hexane and added to a silica gel column prepared from 0.5 g of Kiselgel 60, 70 to 230 mesh (Merck), stored in 130°C, and eluted with 5 mL of n-hexane. The hexane was evaporated to 100 ,uL by using a stream of nitrogen, and 2 pLL of this solution was used for the analysis of PCPYs. The PCDDs and PCDFs were separated from other polychlorinated impurities by using an Alox-column (1.0 g basic alox activity 1, Woelm, ICN West Germany, stored at 190°C). The first fraction consisting of 1 mL n-hexane and 4 to 6 mL of n-hexane: methylene chloride (98:2) was dis-   charged. The second fraction, 10 mL n-hexane: methylene chloride (1:1) was collected and evaporated until dryness in a stream of nitrogen. The residue, dissolved in a known amount of solvent (100 ,uL), was used for the GCIMS analyses.

Isomer-Specific HRGC/MS Analyses
We used a 25 m SE 54 capillary column for screening and a 60 m SP 2330 for the isomer specific analyses. The GC conditions are given in Table 3 (Table 4). The capillary column was attached directly into the ion source of the MS instrument.
In our laboratory we have access to all tetra-, penta-, hexa-, heptaand octa-isomers in the PCDD and PCDF series. The synthesis of these standards has been described elsewhere (2)(3)(4). The separation ofthese isomers using the SP 2330 column is illustrated in Figure 4 for the PCDDs and Figure 5 for the PCDFs.

Results
Transformer Fire, Binghamton, NY, USA On February 5, 1981, a fire in the State Office Building in Binghamton, NY, USA caused a transformer to rupture, releasing soot into the whole building. The external energy source was the electricity, apparently more than 30 min elapsed until the electrical currect to the transformer was switched off. The dielectric fluid in the transformer consisted of a mixture of PCBs (65%) and chlorinated benzenes (35%).
A soot sample was analyzed and found to be highly contaminated by PCDFs and PCDDs. The total level of PCDFs was 2160 ,ug/g, and 20 pug/g of PCDDs were also reported (5). The detailed isomer-specific analysis ofthis soot sample is shown in Figure 6 (PCDFs) and Figure  7 (PCDDs). In this soot sample we also reported on polychlorinated biphenylenes (PCBPs) and pyrenes (PCPYs) (5).

Capacitor Explosion, Stockholm, Sweden
On August 25, 1981, an explosion took place in a capacitor bank (11 kV) in an electrical power station at Danviken, Stockhom, Sweden. The exposion was caused by an electrical failure in the capacitor bank and was so violent that it caused the wall to rupture. The explosion was also associated with external arcing.
In a wipe sample collected about 1 m from the exploded capacitor, we have earlier reported that the major products were a series of chlorinated biphenylenes (PCBPs); the total levels of PCBPs was reported to be in the range of 25,000 to 30,000 ng/m2. The existence of PCBPs was confirmed using the modern MS/MS technique. In addition to PCBPs and PCDFs we also found a series of polychlorinated pyrenes (PCPYs) (5). The levels of PCDFs are reported in Table 5, and the mass frangmentograms of the tetra-, penta-and hexa-CDFs are shown in the upper curves of Figures 8, 9 and 10. Capacitor Fire, Skovde, Sweden On March 19, 1982, a violent fire broke out in a capacitor bank (400 V) serving a high-frequency oven in a casting line of a metal treatment factory (Volvo) in Skovde, Sweden. The dielectric fluid in the capacitors consisted of mineral oil or PCB, and the two types of capacitors were mixed at random. The fire started in a capacitor containing mineral oil and spread to others also filled with mineral oil. The PCB-filled capacitors were strongly heated. A time period of 2 hr elapsed until the fire was completely extinguished. The capacitor room situated in the basement of the foundry building was burned out. Inspection revealed that the copper bars for electricity in the ceiling was partly melted (mp, 1080°C). The smoke spread in the building in an area of 60 x 30 m, and the building was efficiently ventilated through the whole time period.
The capacitor bank consisted of 40 units and 21 of these were filled with PCB, 5 kg each. An inspection after the fire showed that 12 of these capacitors were open and 9 of them were still sealed. Four wipes were collected at the following places: on the floor in the capacitor bank; on the floor in the capacitor room close to the bank; on the wall in the capacitor room, 2 m from the bank 3 m above the floor; on a bench in the ground floor, 10 m from the oven. The quantitative results are collected in Table 5. These data show that high levels of PCDFs (> 100 ng/m2) could only be found in an area close to the fire. At a distance of 10 m from the fire the levels were found to be about 100 times lower. The low value found in sample number 3 can be explained by the high temperature at this sampling location (the copper bars were partly melted here). The mass fragmentogram of the tetra-CDFs is given in Figure 11. The 2,3,7,8-and 2,3,6,7-isomers are major peaks eluting at the end of the chromatogram.
The major products in these samples are the polychlorinated pyrenes (Fig. 12). The di-CPYs are the dominating compounds. However, final proof cannot be obtained due to lack of synthetic standards.
Capacitor Fire, Surahammar, Sweden On September 23, 1982, an oven at the Surahammar steel mill was ruptured by an explosion. The molten steel, 5 tons, 1500°C, ran down into the basement, passed through two safety doors (steel) and ignited a bank of capacitors consisting of 500 units filled with PCB (approx. 250 units) and mineral oil. A violent fire took place, and the whole building (140 m x 70 m x 25 m) was filled by smoke.  Immediately after the fire wipe samples were collected and analyzed ( Table 6). The highest levels were found in the capacitor room and in the higher parts of the building. On the floor the levels were found to be much lower. The building was cleaned by vacuum cleaning and high pressure washing. Due to dusting problems the washing had to be repeated. The whole operation took 3 months. Samples collected after the cleaning were analyzed in the same way and they showed a dramatic decrease in the PCDF levels ( Table 6). The distribution of the individual tetra-, penta-and hexa-CDF isomers is illustrated by the fragmentograms in Figure 13.

Imatra, Finland
Early in the morning August 2, 1982, a bank of capacitors in a paper mill outside Imatra in SE Finland was destroyed by explosions and fire. A wipe test was analyzed; the levels are given in Table 5, and the fragmentogram of the tetra-CDFs in Figure 11 (lower curve). It is clear that the isomeric pattern in this sample is

Hallstahammar, Sweden
On November 8, 1982 a capacitor exploded in a foundry in Bulten-Kanthal in Hallstahammar, Sweden. The capacitor (Fig. 14) was placed close to a melting oven in the ground floor, and the fumes spread in the whole building (40 m x 15 m x 5 m). Wipe tests were collected before and after the cleaning operations; vacuum cleaning and high pressure washing. The cleaning operations were completed in one week. The quantitative results are collected in Table 7, and the fragmentograms are given in the middle curves in Figures 8-10.

Railway Locomotive, Sweden
During the fall-winter period 1982/83 another capacitor exploded in a railway locomotive in S Sweden. The accident was not reported immediately and the samples were collected in February 1983. The results are given in Table 8   the debarking line in a saw-mill in Kisa, Sweden. The explosion was associated with external arcing and fuming and the dielectric fluid splashed around. Three days later samples were taken and analyzed (Table 9).

Used Capacitors
We have opened some used and swollen capacitors, collected and analyzed the drained PCB. The levels of PCDFs was approximately 2 ppm, mainly tetra-CDFs, which is similar to the levels we found for the new and unused PCB (Prodelec 30 10). Examination of the paper layer revealed dark spots, probably caused by electrical arcing. The PCB around these dark spots was extracted and analyzed. The levels were similar to the levels found in the drained PCB or new PCB.

Sparking Experiments in PCB
No increase in PCDFs (ca. 2 ppm) could be found after up to 7 sparks (18 kV, 2 ,uF) in pure PCB. No PCBPs could be found; the ratio PCDFs:PCBPs was greater than 100:1. In another experiment 1% of 1-epoxyethyl-3,4-epoxycyclohexane (Union Carbide ERL 42 06) was added before the sparking (4 sparks, 15 kV, 2.5 ,uF). In this experiment, the PCDF levels increased (Fig. 15). In general the level of PCDFs increased ten times, but for a few isomers the increase was 30-fold; for others no increase was found. The ratio PCDFs:PCBPs was greater than 100:1.