Pyrolysis products of PCBs.

Model compound studies which were previously done for impurities and environmental residues of chlorophenols and for wastes of chlorination processes were extended to the impurities and pyrolysis products of polychlorinated biphenyls (PCBs). Model compounds were commercial products or synthesized and their structures proven by spectroscopic methods. These models were used as analytical reference substances in GC/ECD and GC/MS studies of the pyrolyzed PCB samples. In addition to previously known neutral components like polychlorinated dibenzofurans (PCDFs), chlorophenolic substances, especially polychlorophenols (PCPs) and polychlorinated biphenylols (PCB-OHs) were observed as major pyrolysis products of PCBs. Capacitor fires are suggested to produce in many cases chlorophenols which are major toxic hazards to people.


Introduction
Our laboratory has been working on structural and environmental chemistry of polyhalogenated aromatic compounds since 1973, when we made the first observations of the occurrence of chlorophenols in aquatic food chain samples (1). The first model compound studies, development of trace analysis methods and analyses of environmental samples were done for chlorophenols, catechols, cresols, guaiacols and their derivatives. These compounds were considered important wastes from wood preservation and chlorination processes like chlorobleaching of pulp, chlorodisinfection of water and later also as combustion residues (2-38). 2,3,7,8-TCDD as an important possible co-contaminant was added to our studies in 1976 (38)(39)(40)(41).
Other polyhalogenated aromatic compounds containing two or more benzene rings were included to our structure analyses in 1978 in order to monitor impurities of the wood preservative chlorophenol formulations in their preparation and waste treatment (42). Especially, polychlorinated phenyoxyphenols (PCPP) were synthesized by us to complete the series of these important chlorophenol impurity models (43), which had been started by Nilsson in his Ph.D. thesis (44 -46). Recently, *Department of Chemistry, University of Jyvaskyla, Kyllikinkatu 1-3, SF-40100 Jyvaskyla 10, Finland. t2Section Chemie, Karl Marx Universitat, Leipzig, DDR. **Research Centre of Defence Forces, SF-34111 Lakiala, Finland. tAuthor to whom requests should be sent.
we have studied dimeric polyhalogenated aromatic (PHA) impurities in spent chlorobleaching liquors and pyrolysis products of PCBs by gas chromatographic and mass spectrometric analyses and by comparisons with available model compounds. Observations on the pyrolysis products of PCBs are described in this paper.

Components Studied
Structures and commonly used abbreviations of the names of neutral components and pyrolysis products of PCBs are shown in Figure 1 and those of the possible phenolic pyrolysis products in Figure 2.
Structure analyses of the toxic pyrolysis products of PCBs have been intensively made since chick edema disease was experienced in the US broiler industry in 1957 and caused deaths of millions of chickens (47). A major toxic product was determined 1969 by crystallography to be 1,2,3,7,8,9-hexa CDD (48). PCDFs and PCNs were observed in commercial PCBs in 1970 (49). In 1968 more than 1000 persons in Japan were poisoned by a rice oil (Yusho oil) contaminated by PCBs used as a heat transfer medium, and PCDFs were found as main toxic factors (50). Polychlorinated terphenyls (PCTs) which had been found in PCB were also identified as environmental poisons in 1977 (51). The importance of combustion as source of such toxic and persistent materials was exposed in 1977, when PCBs among other chlorohydrocarbons (52) and also PCDDs and PCDFs Formation of PCDFs and other neutral PCB pyrolysis products in fire accidents has been monitored (62)(63)(64). In addition to product types mentioned above, PCBP and polychlorinated pyrenes (PCBPy) (Fig. 1) have been reported to be formed in PCB fires (64), and 2,3,6,7-TCBP has also been studied as an TCDD-type superpoison (65).

Quartz Tube Pyrolysis
The Hamilton multipurpose sampling system was used. A sample of the PCB mixture (Aroclor 1254) was weighed into a 5 cm quartz tube of 0.8 mm internal diameter which was then stoppered with quartz wool and put into a quartz pyrolysis tube of 4 mm internal diameter. An air stream (40 mL/min) was started, and pyrolysis was done in the 20 cm heating zone at 500, 600 and 700°C followed by heating in the 10 cm tube zone at 300°C. The adsorption tube was then filled with 1 g of XAD7/XAD4 (1:1) mixture. Pyrolysis time was 3 seconds. Three successive samples of about 30 mg were pyrolyzed and all products were trapped in the same adsorption tube.

Fractionation of the Pyrolysis Products
Pyrolysis products were desorbed from the XAD mixture by eluting with 20 mL of an acetone-hexane 1:1 (v/v) mixture. The evaporated eluate was dissolved in 25 mL of hexane and shaken in a separatory funnel with 50 mL of 0.1 M K2CO3 water solution. The separated water phase was washed five times with 20 mL of hexane, and the organic phases were combined to form the neutral product fraction.
The water phase was shaken with 1 mL of acetic acid anhydride and allowed to stand for 5 min. The acetate derivatives of phenols thus formed were taken into 20 mL of hexane by shaking. The hexane solution was used for GC/MS and GC/ECD analyses of phenolic fraction. Extra solvent was evaporated from the neutral product fraction, and the residual 2 mL solution was shaken with 2 mL of concentrated sulfuric acid and then fractionated in an alumina microcolumn following the procedure of Buser (68) by eluting PCBs with 2% methylene chloride in hexane (fraction I) and PCDFs and PCDDs with 50% methylene chloride in hexane (fraction II). The latter fraction was investigated by GC/MS and GC/ ECD.

Analysis of the Capacitor Fire Fumes
Samples of carbon filter containing fire fumes were weighed and Soxhlet-extracted with toluene for 24 hr. The phenol fraction was taken in 0.2 N NaOH by two funnel extractions, washed with hexane and neutralized with dilute sulfuric acid to pH 7. Potassium carbonate was added to 0.1 M solution, and the procedure continued with acetic anhydride as in the case of pyrolysis products (see above). The hexane, in addition to PCB and other neutral impurities, removed 3-, 4-and 5-chlorophenols which made quantitation a problem. (GC/MS in Figure 8 is from a sample washed eight times, and thus only a small part of the original 3CP, 4CP and 5CP remains in the phenolic fraction.) The combined toluene and hexane solutions were evaporated and handled for analysis of neutral components as in the case of the pyrolysis products (see above).

GC/MS Analyses
A Finnigan model MAT 212 system was used. The GC column was a 20 m long, 0.3 mm diameter quartz capillary coated with SE-54 to film thickness 0.4 ,um. Both chlorophenol acetate and neutral fraction (fraction II) were injected with 1:10 split at 260°C. Helium was used as carrier gas 1 mL/min. Temperature program of 600 + 10°C/min to 260°C was used. An open split and quartz capillary connection led to the ion source having a temperature of 200°C where 70 V electron impact was applied. Continuous scans at the mass number range 50-550 were run at a speed of 2.3 sec per full scan; mass spectra were collected in data storage using 1000 x resolution (10% valley).

GCIECD Analyses
A Micromat HRGC 412 two-column system was used. Both columns were 25 m long, 0.3 mm diameter quartz capillaries. One column was coated with SE-54 and the other with OV-1701 to a film thickness of 0.25 ,um. Injection with 1:10 split was applied at 275°C. The carrier gas was helium (1 mL/min) and the makeup gas to the 'Ni EC detectors was argon-methane, 95:5 (v/v) at a rate of 18 mL/min. Temperature programs used were for chlorophenol acetate fraction 100°C +4°C/min to 250°C and for the neutral fraction (fraction II), 60°C + 10°C/min to 280°C.

Identification
Structures of the components were deduced primarily from the mass chromatograms of selected ions precalculated for each type (Figs. 1 and 2) and each number (P) of chlorine atoms. After detection of a peak for the selected ion, the corresponding mass spectrum was reconstructed and the structure ofthe component verified. If model compounds of the type were available, they were chromatographed by GC/MS and GC/ECD exactly under the same conditions as actual samples. Retention times at each column and MS comparison gave a firm identification of the structure of the component. The minor components (which gave mass chromatogram peaks of too low intensity of MIZ of their ions to be recognized) could in some cases be detected from the total mass spectra of major components as clearly resolved extra ion clusters ("impurity peaks"). and 8CDD) in our laboratory were used as external standards and in some cases weighed and added to the analysis mixtures to obtain comparison of the GC/MS and GC/ECD peak intensities. While, however, identical models were absent for most dimeric components in the complex product mixtures, response of an "average model compound" was taken as quantitation standard for each structure type. 2,4,5-Tribromophenol was used as response standard compound for both monomeric and dimeric chlorophenols, and 1,2,3,4-tetrachlorodibenzo-p-dioxin was used for neutral dimers. Due to difficulties in fractionation and cleanup operations, the absolute amounts could, however, only be estimated to an accuracy of one order of magnitude. Relative amounts within compounds types could be obtained more precisely from the present data. From reconstructed mass chromatograms the areas of the M-42 ions of the phenolic components and the M-ions of the neutral components were compared. In this way, ratios of 74 product components within PCP and PCB-OH groups listed in Table 1 were calculated, and the results are presented in Table 2. Also ratios of the main neutral pyrolysis products within groups PCDF and PCDD were calculated, and the results are collected in Table 3. Table 3. Relative amounts of the major neutral pyrolysis products formed from PCB (Aroclor 1254) in quartz tube pyrolysis in air stream at 500, 600 and 700°C and in carbon filter from a capacitor fire.

Results and Discussion
Products of the first experiment, open flame (Bunsen) pyrolysis at 300-400°C were analyzed by GC/ECD using model compounds and only the monomeric chlorophenols (PCPs) were sufficiently well resolved to be identified and quantified. As a result, the following production of PCP groups based on the original amount of Aroclor 1254 was measured: 1CP, 4 ppm; 2CP, 3 ppm; 3CP, 2 ppm. GCIMS also detected PCB-OHs; the estimated amounts were: 2BC-OH, 0.1 ppm; 3CB-OH, 0.2 ppm; 4CBOH, 0.2 ppm; 5CB-OH, 0.05 ppm; 6CB-OH, 0.01 ppm.
GCIMS and GC/ECD chromatograms of the phenolic product fractions from pyrolyses and capacitor fires are illustrated in Figures 3-8; qualitative analysis (detection and identification) results of this fraction are given in Table 1. Reconstructed MIS gave ratios of the main components which are collected in Table 2. Absolute amounts could be estimated reasonably only from the ifiter sample when the phenol fraction was washed eight times with hexane: 20.6 ppm PCP and 12.8 ppm PCB-OH were found in the filter. The real amount of PCP must have been much higher, but hexane has removed most of PCP. 2,4,6-Trichlorophenol and 2,3,4,6-tetrachlorophenol (structure nos. 10 and 17) were especially abundant in nonwashed filter sample.
Neutral fractions I were only preliminarily investi-    Table 1.  Still very much lower amounts of ICPy and 2CPy were detected in neutral fraction I of the pyrolysis at 600°C. GCIMS chromatograms ofthe neutral fraction II were similar to those published in earlier papers (55)(56)(57)(58)64). in reconstructed mass chromatograms from fraction II run with selected ions, however, large peaks with high SIN ratio were obtained from PCDFs (major product components) only. On reconstructing total mass spectra, some PCDD, PCN and PCPy component types were also detected. Main neutral dimer structure types found in pyrolyses were (in order of abundance) 3CDF, 4CDF, 2CDF, 4CDD and 5CDD. 1CDF and 6CDF were detected as minor components. In pyrolysis of Aroclor 1254 at 6000C, 0.16% conversion to PCDF and PCDD was estimated, which is somewhat lower but ofthe same order of magnitude as reported by Buser, Bosshardt and Rappe (56).
Ratios of the main neutral components (from reconstructed MIS) are collected in Table 3. The absolute amounts of PCDF in the Enso filter were also estimated to be 15.8 ppm in carbon. Consequently, chlorinated phenols appear to be an order of magnitude more abundant in capacitor fire products studied here than PCDFs.
Our present finding of high amounts of PCP and PCB-OH in PCB pyrolysis and capacitor fire products and abundant formation of only 1-3CDF in the latter calls for new toxicological evaluation of the risk components in certain accidents.
We are grateful to Simo Lotjonen, Lic.Phil., University of Kuopio and to Heikki Pyysalo, VTT, Helsinki for the PCB-OH and PCB(OH) model compounds. Financial support of Maj and Tor Nessling Foundation and Enso Gutzeit Co. is gratefully acknowledged.