Polycyclic aromatic hydrocarbons in the work atmosphere I. Determination in an aluminum reduction plant

,and P. E. Polycyclic aromatic hydro carbOillS in the wor'k ,atmosphere: 1. Determinaltion in an, alUlmtnum reduction plant. Scand. j. work environ. & health 4 (1978) 212-223. The content of polycyclic aroma tic hydrocarbons (PAH) in a Sederberg prebaked aluminum reduction plant and an anode plant was studied. Both stationary and personal sampling were used, and the samples were analyred by glass capillary galS chroma,tography. In many samples more than 30 PAH and heterocyclic compounds were identified. Compari son of tberesullts ~rom Ithe different planJts revealed 1lha,t the atmospheric PAH content was signHiicantly higher in the Seder,berg plant than in the others. How ever, the relaillive dist,ribullion of PAH components (PAH profile) was relatively constant :in ,the prebaked and the anode lba1cing plant, ,and different in the Seder berg plan'1;. As a consequence of the va,riaiion [n the po11ullion 'level among the dUferent job caJtegories, personal sampling showed a broader -range of PAH con cenrt;rations than the staillionary method.

It is well known that polycyclic aromatic hydrocarbons (PAR) may be formed by pYl'Olysis and inoomplete combustion of ooal, tar, oil, and other organic material (8). In an industrialized society there exists a large number of anthropogenic sources of PAR. Much interest has been focused on analyzing and characterizing these sources and the distribution of PAR, both in the outdoor environment and in 1  212 work atmospheres. The reason for this is basically that animal experiments have shown that some PAR have carcinogenic properties, and they are suspected to be human carcinogens as well (8). It has also been shown that smokers or persons br,eathing air containing PAR dev,elop lung cancer more often than the general population (8). An epidemiologic investigation among employees in a coke plan t rev,ealed a 2.5 to 10 times higher cancer fr,equen<:y than among the control group (7). Most ana'lyses of PAR in the work environment have, :however, been limited to the determination of "benzene soluble material" (9) or to the determination of one single component, such as benzo(a)pyrene, while complete information on the PAR content in atmospheres is more scarce.
In parallel with the increasing concern for occupational exposure to hazardous compounds, there is al90 ,a need for moredetailed knowledge of the oomposition and pollution levels in various work atmospheres. The mixture of P AH formed by high temperature reactions may be of a very oomplex nature and may consist of a large number of carcinogens and noncarcinogens. Recent results have demonstrated that PAH compounds also exhibit synergistic effects (10). Detailed knowledge of the composition of the PAH compounds in a work atmosphere is 1iherefore essential for an estimation of possible hazardous effects.
In 'connection with a research program on the occurrence and analysis of P AH in work atmospheres, we have previously reported ona method for 'analyzing P AH (2) and the PAH ,content of airborne particulate .pollutants in a S0derberg paste plant (4). In this paper we present the results of an investigation of the PAH content both in air and in airborne particulate matter in a Norwegian primary aluminum smelting plant (3). The aluminum plant has both v,ertical pin S0derberg a'lld closed prebaked anode cells. In addition there ar,e an mode paste and an anode ba!king plant. The work atmospheres in these plants have been investigated, and the results are compared. As a means of comparison, the ooncept of P AH profiles has also been utilized.
MATERIAL AND METHODS oration of etha:nol. (Under lIlormaloondi-tiQ11s, 'about 20 ml lof ethanol!h evapoflate from :the bottles.) An Edwards ECB 1 vacuUiIIl pump drew the air through the equipment with a flow rate of about 1 m 3 /h. The sampling volume was measured with a dry gas meter.
In !personal sampling, two types of devices were used. The first one consisted ofa 27 mm 0 Acropore AN 800 filter, connected to an Casella pump. The filter wa,s placed in the breathing zone of the workers, and the sampling rate was about 2 lImin. Wi'th this method we collected only partieulate matter.
The other type of personal sampling device is, in .prin~iple, equal to the stationary device. By means of a batterydriven pump (flow rate 'V 2 lImin), air was dr,awn through a ,37 mm 0 Acropore AN 800 filter and then bubbled through two leakproof aboorption bottles with 7 ml of ethanol in each. The botHes wel'e oooled with dry ice. The sampling unit weighed 3 'kg and was carried around in a little back:pa~k. The sampling time varied hetween 2 and 6 h.

Sampling sites
Air samples were oollected at a number of .places in the anode plant (production and prebaking of anodes and production of anode parts), the vertical pin S0derberg plant, and the closed prebak:ed anode  plant. The filter house was placed about 1 m abov,e the floor and the sampling time varied between 3 and 5 h. The personal samplers were carried by workers with different jobs, as described later.

Analytical procedure
The analytical method has been reported previously (2) and will only be described briefly now. The internal standards (,8,,8'-binaphtyl and m-quaterphenyl) were added to the -air filter soamples, which were subsequently extractoo in a Soxhlet apparatus for 8 h with 50 ml of cyclohexane. The cydohexane solutions were extracted with 50 ml of N,N-dimethylformamide (DMF) : water (9: 1), which transfer PAH 214 and more polar compounds to the DMF phase. '.Dheextraction was repeated with 25 ml of the DMF/water solutton. To the combined DMF/water phases were added 100 ml of wa,ter, and PAH was reextracted with 50 ml of cyc1ohexa,ne. The extraction was repeated with 25 ml of cycloihexane. The combined cyclohexane phases were dried by Na2S04 and concen.trated to about 10 ml in a modified Kudema-Danish apparatus. Further concentration was a,chiev,ed in a centrifuge tube on an aluminum heating block at 30°C and a g.entle stream of highly purified nitrogen (> 99.999 %).
The separation of the extract was performed on ·a Carlo Erba gas chromatograp'h model Fractovap 2100 A, equipped with a glass capillary column and flame tonization detector (FID). The sample was intr.oduced .on the gas chromatographic column with the technique of splitless inj,ection (5). The chromatographic condihons are given in table 1.
The PAH components were identified in a oomparison of the retention times with those of a set of P AH standards CliIld with chmmatograms ir,om previous studi,es in which mass spectrometric iden tifioation was used (4). The oompounds were quan.tified by electronic intJegr,ahon of the peak areas (Minigrator, Spectra-physics) alIld by a comparison. of <these areas wttlh the areas of the internal standards.
The precIslon of the method measured as one standard deviation has been reported to be better than 14 Ofo (average of 15 compounds is 4.8 Ofo), and the goodness of the method is 11.6 Ofo (2).

RESULTS
In the course of this work, a number of air filter samples and absorpti'on solutions were analyzed. In the foUowing, the results for those analyzed by gas chromatography are presented separately for each  of the three plants. In addition, a few results from a high-pressure liquid chromatographic (HPLC) analysis are also given for comparison.

Anode plant
Air samples were collected by stationary sampling at different sites in the plant. Four of these sites were located aboV'e the baking ovens (samples 1-4), while the fifth was above the anode compressor (sample 5). The results 'are giv,en in table 2, both f.or the :partkular matter and for 216 the absorption solutions. In addition, a number of air filter samples from personal samplers hav,e been analyzed. The results of four of these are given in table 3. All the samples represent different work operations, such as coke packer (sample 6), pitch bin worker (sample 7), hydraulic press operator (sample 8) and pitch dust sweeper (sample 9).

Prebaked plant
Stati~mary sampling was also used in the prebaked plant. Results from analyses of  The results of the analysis of stationary samples are given in table 6 (samples 16-25). Personal sampling was used in this plant as well, and the results are given in table 7. These samples in.clude·operations such as tapper (sample 26) and pin setter (samples 27 and 28). In fig. 1 a chromatogram of PAR from particulate matter in the atmosphere in this plant is giv·en. Fig.  2 shows a chromatogram of the volatile PAR fraction from the same sampling. As revealed by this chromatogram, PAR up to pyrene is likely to pass through the filter.

DISCUSSION
As shown in the tables, more than 30 different PAR oompounds were identified in the samples f!"Om the work atmospheres. Carcinogenic ,activities as revealed in animal experiments (8) have been assigned to a few of these, i.e., benz(a)antracene, benzo(c)phenanthrene, benzo(b&k)fluoranthene, benzo(a)pyrene and dibenzopyrene. Previously, the main attention to potential carcinogenic compounds in work atmospheres has been focused on the occurrence of benzo(a)pyrene (BaP). BaP has also been suggested as an indicator of the PAR level in work .atmospheres oontaining tar and piotch volatiles (11). The present study demonstrates the existence of other potentially hazardous materials as well.
The total concentration of PAR also varied within the three plants. The results from the stationary sampling in the Sooerberg plant showed an average value of 0.1 mg/m 3 for PAR on particulates and 1 mglm 3 for volatile PAR. The personal sampling filters showed values for PAR on particulate matter from 0.003 to 2.8 mg/m 3 . For the prebaked plant, the values obtained from the gas chromatographic analysis were quite low. On the average the stationary sampling gave values of parti-culate PAR of the order of 0.001 mg/m 3 and volatile PAR of about 0.07 mg/m s , which is in the region of 1-10 % of the amount in the Sooerberg plant. The personal sampling filters also showed quite low values. For PAR in particulate material it was determined to be about 0.001 mg/m 3 . It is, however, necessary to point out that these results are not representative of the PAR concentration in the atmosphere of the prebaked plant, since only those filters with high amounts of PAR have been analyzed by gas chromatography. The results from a RPLC analysis of filters with less PAR should also be taken into consideration when the occupational      hazard of PAR in this envi:mnment is evaluated.
In the anode baking plant (samples 1 to 4), the average concentrations of PAR (stationary sampling) were determined to be 0.004 mg/m s in particulate materials. A maximum value of 0.07 mg/m s was determined in a sample from the atmosphere above the anode press (sample 5). The personal samplers in this plant gave concentrations of 0.01-0.4 mg/m s .
In the anode paste plant there are two jobs with particularly high exposure risk, i.'e., coke packing and working in the pitch bin. The coke packer had a concentration of 0.037 mg/m s PAR. In the pitch bin, both 222 stationary and personal sampling gave PAR concentrations on the order of 0.4 mg/ m S in the particulate matter.
A recent paper by Jackson et al. (6) reported on BaP lev'els in 'a coke oven ploot of 1.8 fhg/ms. In our study we observed values of 0.02-0.05 fhg/m3 in the prebaked potroom, 0.7-9.0 fhg/ms in the Sederberg potroom, and 0.03---0.3 !A'g/m s in the anode plant, all values for stationary sampling. As expected, a lower PAR content was found in the potroom atmosphere of a prebaked plan t in comparison to that of the vertical pin S0derberg plant A schematic listing of 'all the samples of PAR in particulate matter in the persona,l and stati,onary sampling as a function of sites is given in fig. 3. As expected, the personal samples show more scattered values than do :the sta:tionary ones, probably due to the fact that the oven control is performed only occasionally, and henoe the expo,sure varies, and in addition different job situati:ons iJnvolve dtfferent levels of exposure.
In order to compare the analytical results from the different work atmospheres, we utilized the concept of PAR p1"ofiles (4). We constructed the PAR profi'le by plotting the rela'tive distribution of selec!r ed PAR components in the omer of elution from the gas chromatographs. It has previously been postulated that the PAR pr,ofile is a characteristic of the process involved, i.e., the basic composition of the organk material, redox conditions, and reaction temperature. In fig. 4, the PAR profiles of the particulate matter in the work atmospher,es in the th1"ee pl;alllts are depicted. As revealed in the figur,e the PAR profiles of the prebaked and anode baking plants are relatively parallel. Since these processes essentially use the same carbon material to approximately the same temperatur,e, the results support the idea that the profile is a characteristic of the PAR source. The Sederber,g plalllt, whioh has a different process, also gave a different PAR profile ( fig. 4). In this case, a significant higher fradion of the PAR belonged to the high~hoiling compounds than was shown for the anode prebaked plants.
The PAR profile is important for several reasons. If the profile is fairly constant with time, the monitoring of PAH can be simplified, since indicator substances [like benzo(a)pyrene] can be used. Determination of the total amount of P AH is also more meaningful when the profile is known. Finally, there are some indieations that the ,carcinogenk acttvity is a function of molecular size (1) and that the PAH profile may be used for the evaluation of potenmal PAH hazards in work atmospheres.
Additional studies of PAH in different branches of the electrochemical and other industries are in progress.

ACKNOWLEDGMENT
This wo,rk is part of a joint Norwegian project on "Polycyclic Aromatic Hydrocarbons :iJn the Wor.kimg Environment" between the Institute of Occupational Health, The Engineering Research Foundation at the Technical University of Norway, and the Oentr,al Ins1Ji'tute for Industrial Research.
The authors wish to express their sincere thanks to B. H{)vdal, K. Halgard, and B. Olufsen for their assistance in <JoHecting and analyzing the samples. A/S Ardal og Sunndal Verk is thanked for their support during the sampling part of this work.
Financial support from The Royal Norwegian Council for Scientific and Industrial Resea1"ch IlJIlder oontract B 1551.4791 is gratefully acknowledged.
Received for publication: 28 November 1977