Thermal oxidation of polycyclic aromatic hydrocarbons adsorbed on alumina.

The reevaporation of polycyclic aromatic hydrocarbons (PAH) from secondary alumina (alumina used in a fluidized bed for pot gas cleaning) used in the production of aluminum in aluminum reduction plants has been studied. The secondary alumina contains around 100 ppm of PAH when introduced to the pots. The results of this study indicate that only a minor part of the adsorbed PAH is reevaporated at the pot temperature of 300-400 degrees C and that the use of secondary alumina has little effect on the PAH concentration in the workplace atmosphere of the aluminum reduction plants.

The occurrence of polycyclic aromatic hydrocarbons (PAH) in the work atmosphere of aluminum reduction plants (3, 4, 5, 10,12) and in the emissions from such plants (1) is well established. The PAH is formed during the electrolysis process by pyrolysis of the coal tar pitch and coke which constitute the electrodes used in the Soderberg process.
The emission of fluoride has been considered the most serious environmental problem caused by aluminum smelting. Pot gas cleaning through a fluidized bed of alumina (primary alumina) has greatly reduced the fluoride emission. The alumina in the cleaning system also adsorbs some of the emitted PAH. After passing through the cleaning system, the alumina (now called secondary alumina) is added to the electrolysis pots.
The analysis of several extracts from secondary alumina has shown it to contain PAH in concentrations around 100 ppm. The recognition of such relatively high concentrations of PAH in the alumina has caused some concern about the possible impact of the alumina-bound PAH on the total concentration of PAH in the workplace atmosphere. As the alumina, when added to the pots, is heated to 300-400°C, it was feared that the adsorbed PAH might evaporate and increase the ambient concentrations of these unwanted compounds in the workplace atmosphere.
The present study was initiated to determine to what extent such evaporation takes place.

Alumina
The alumina used in this study was primary and secondary alumina from a vertical stud Soderberg plant with 100 070 alumina feeding (all primary alumina passes through the dry-cleaning system). The specific surface of the alumina was in the range of 45-55 m 2/g. The alumina is normally characterized with respect to several inorganic parameters before and after its use in the cleaning system. Because of a rapid, largevolume alumina stream through the dry-cleaning system, the concentrations of pollutants in the secondary alumina is relatively low. The concentrations of sulfur, fluoride, and PAH increase from below detection limits in the primary alumina to about 0.15 0J0 sulfur, 0.8 070 fluoride, and 100 ppm PAH in the secondary alumina with a concomitant decrease of the pH from approximately 10 to about 6.
Heating of the alumina Unless otherwise stated, 10g of secondary alumina was heated from room temperature to 360°C in 30 min in order to mimic the heating process that occurs in the smelting pots. The flue gases were collected with the equipment shown in figure I. In a separate experiment the reaction products were 'sampled for gas chromatographic-mass spectrometric analysis only, and the equipment was then simplified by the removal of the ethanol impingers and the inclusion of a glass fiber filter followed by an XAD-2 adsorbent.

Analysis of polycyclic aromatic hydrocarbons
Internal standards for PAH analysis were added to the alumina, which was subsequently extracted with etha- nol by ultrasonication for 15 min. The ethanol extracts were shaken with cyclohexane, and the cyclohexane extracts were thereafter extracted with dimethylformamide: water (9: 1) as described by Bjerseth (2) with the addition of a clean-up step consisting of passing a concentrated extract through a column of silica and eluting the column with cyclohexane. The concentrated eluate was analyzed on a Carlo-Erba gas chromatograph fitted with a glass capillary column and a flame ionization detector.
For some of the studies a simplified procedure estimating the total amount of PAH by use of thin-layer chromatography (7) was used instead of an analysis of the individual compounds.

Radioactivity counting
In a study of the mass balance, some of the alumina was spiked with 14C-Iabeled benzo-(a)pyrene (BaP). The spiking was achieved by suspending the alumina in pentane in which [Ujl 4C-BaP [21.7 mCi/mmol (868 MBq), the Radiochemical Centre, Amersham] was dissolved and letting the pentane evaporate as the mixture was stirred.
Various extracts of the spiked alumina were counted for radioactivity by liquid scintillation counting in a Tri-Carb spectrometer, model 3003. Quench corrections were performed by use of the channels-ratio method.
The remaining activity on the extracted alumina was measured by suspending the alumina directly in the scintillation liquid, and the number of detectable pulses per minute was calculated by use of the internal standard method for quench corrections.

Gas chromatographic-mass spectrometric analysis
After the secondary alumina was heated, the filter used for collecting flue gases was extracted in a soxhlet apparatus with acetone for 24 h. The solvent of the extract was then changed to methylene chloride and analyzed by gas chromatography-mass spectrometry. The system consisted of a Finnigan model 4021 quadrupole mass spectrometer equipped with a standard 440 electron impact source. Ionization was accomplished with a 70-eV beam of electrons generated from a heated rhenium filament with an emission current of 0.25 rnA. The source temperature was 250°C. The electron multiplier voltage was 1 700 V.
Samples were introduced by means of a Finnigan 9610gas chromatograph directly interfaced to the mass spectrometer by the fused silica capillary column. The typical gas chromatographic conditions were as follows: injector temperature 280°C; interface 240°C; carrier gas helium; linear velocity 40 ern S-1 at Ioo°C; column: 30 m x 0.25 mm inner diameter DB-5 (J & W Scientific), film thickness 0.25 /lm; column temperature program from 100 to 300°C at 5°C mirr", after initial temperature had been held for 3 min.
Masses from 40 to 300 mass units were scanned every second. The ion data were acquired with an INCaS 2100 data system.
An acetone extract of the unheated secondary alumina was also analyzed by gas chromatography-mass spectrometry for comparison.

Analyses of extracts of alumina before and after heating and oj the collected flue gases
When extracts of secondary alumina were analyzed for PAH before and after heating, it was found that the extractable PAH decreased rapidly with the temperature (figure 2). The decrease was also found to be dependent on the time of heating, as illustrated in figure 3.
An analysis of the impinger solutions revealed quite unexpectedly that, when the alumina was heated to 340°C or above for a period of 30 min, no PAH could be detected in the impinger solutions. Gas chromatograms of cyclohexane extracts of the impinger solutions revealed a complex pattern of peaks different from that of the unheated secondary alumina. These peaks represented compounds which were mostly more volatile than the PAH present in extracts from unheated alumina. Most of the chromatographic peaks disappeared when the sample was applied on a column of silica gel and eluted with cyclohexane, an indication that they represented more polar compounds than the PAH.
A further characterization of the compounds that could be found in aerosols above the heated secondary alumina was performed by a gas chromatographmas s spectro meter analysis of the extr acts from the aerosol collected on a glass fiber filter in a separate experiment. This analysis demonstrated that some of the most stable PAH compounds may evaporate from the secondary alumina during the heating process, eg, fluoranthene and chr ysene/triphenylene. Mo st of the compounds identified were , howe ver, oxidation products of P AH. The results are given in table I. Figure 4 shows chromatograms of the extract of unheated secondary alumina and of the extract from the sample collected on a filter during the heating of the alumina. The chromatograms showed distinct differences and therefore demonstrated that the vapors evolved during heating have a different constitution than the P AH on the secondary alumina.

Mass balance oj the adsorbed polycyclic aromatic hydrocarbons upon heating of the secondary alumina
The analyses had given qualitative information that the P AH on the alumina had not simply evaporated, but had been subjected to breakdown reactions. The quantitative aspect of the breakdown was studied by a mass balance estimation of the process. This mass balance was performed with secondary alumina spiked with radiolabeled HaP. Aliquots of extracts of heated and unheated alumina, of the alumina itself, of impinger solutions, and of the ethanol extract of the filter (figure I) were counted for radioactivity . The results from two parallel experiments are presented in table 2. As the counting of the impinger solutions revealed unexpectedly low counts, the heated alumina was extracted several times and with acetone extraction following three consecutive ethanol extractions. A counting of dispersed samples of the remaining alumina showed that, whereas nearly all radioactivity could be extracted from the nonheated alumina, 54 % of the total radioactivity added remained unextracted when the alumina had been heated. The total recovery of radioactivity from the heated alumina was about 75 % compared to the nearly complete recovery from the unheated alumina. The loss of about 25 % could possibly be explained by an inefficient counting of the solid samples, but it was more likely caused by very volatile compounds passing through the impinger solutions . As indicated by the counting results for the sodium and barium hydroxide solutions included to trap carbon dioxide, the total breakdown to carbon dioxide and water was nearly negligible.
The radioactivity that was extracted from the heated alumina repre sented at mo st 25 % of the radioactivity stilI associated with the alumina. Gas chromatography of these extracts showed that no HaP could be detected. As the amount of radioactivity in these extracts corresponded to 27 p.g HaP (combined extracts from experiment 2), which is fa r more than the detection limit (100 ng), the extracted activit y must represent reaction products derived from BaP. During the preparation of the extract for PAH analysis, a solvent partitioning into a nonpolar fraction, a P AH fraction , and a more polar fraction was performed (2). When aliquots of concentrates of these fractions were counted for radioactivity, 47 % of the activity was found in the most polar fraction, whereas the nonpolar and the P AH fraction contained 0.3 and 9.6 % , respectively. The remaining 43.1 % was lost, probably through evaporation during the concentration of the fractions, and thus most likely represents volatile compounds.     a Extract from the filters and ethanol wash of the equipment contributed insignificantly and has been omitted from the table. b Two or three consecutive extractions were performed and counted separately. c Impinger solutions from the two ethanol bottles were counted separately.
alumina surface; and (iii) the PAH as such may form coordination complexes with the alumina surface due to alterations in the surface.
In an investigation of which of these theories was the most likely, the experiment using radiolabeled BaP coated on the alumina was repeated once more. This time the alumina was extracted by ultrasonication eight times successively with various solvents . Aliquots from all the extracts were counted for radioactivity. The results are illustrated in figure 5. Dimethylformamide was found to be more efficient than ethanol for the heated alumina, and 42 0,10 of the added activity was recovered from the heated alumina after three successive extractions with ethanol followed by three extractions with dimethylformamide, as opposed to 92 % for the unheated alumina.
No BaP could be detected in any of the extracts from the heated alumina when subjected to PAH analysis. This result was in contrast to results from the extracts of the unheated alumina, which contained 7.6 ppm in the combined ethanol extracts and another 0.7 ppm in the dimethylformamide extracts.

444
Treatment of the heated alumina with hydrogen fluoride released most of the remaining radioactivity. Subsequent extraction of the hydrogen fluoride slurry with cycIohexane and dichloromethane showed that most of the activity was extracted with dichloromethane. No BaP could be detected in any of the extracts.

Discussion
The results from the analysis of the hydrogen fluoride slurry extracts point towards our hypothesis i as the most likely, ie, that the PAH are oxidized to more polar compounds which are bound more firmly to the surface of the alumina. If the radiolabeled BaP had been bound to the alumina surface as part of a larger condensed system or as a coordination complex, it is likely that the treatment with hydrogen fluoride would have produced compounds that were more easily extracted into cycIohexane.

Acknowledgments
Ms MB Stebet and Ms A Osvik are thanked for their skillful technical assistance. Financial support from the Nordic Aluminum Industries Secretariat for Health, Environment and Safety, Oslo, and from the Jahre-Wellcome Foundation is gratefully acknowledged.

References
is used in the smelting pots. The results obtained in this study should be verified in comparisons of P AH concentrations in samples collected from pots operated with primary and secondary alumina.
There is ample evidence that most PAH undergo photooxidation in solution, as pure solids and even when adsorbed to certain solid substrates (13) . The photodecomposition products are mainly ketones and quinones (11). Hydroxy and dihydroxy derivatives are less abundant.
Similar oxidation products may also be formed in the dark when PAH compounds are adsorbed to a solid carrier, but not when the compounds exist in their crystalline form or in solution (11). The rate of the oxidations taking place in the dark depends mainly on the adsorption substrate (II). An adsorptive association of P AH on silica and alumina tends to increase their susceptibility to photooxidative attacks, whereas nonphotochernical degradation of, eg, fluorene has not been observed on these surfaces (II). Adsorption on coal fly ash stabilizes the P AH against photochemical attack, but at the same time nonphotochemical oxidation is permitted to occur (11). Fitch & Smith (6) have found the oxidized PAH to be more readily extracted from the carrier than the parent PAH, an observation that disfavors our hypothesis i.
A theory involving complex formation has been suggested by Fox & Olive (8)

and by Thomas et al (IS).
Fox & Olive suggested that exitation by, eg, light can initiate polymerization pathways which lead to highly condensed systems. Thomas et al (IS) studied the influence of surface area on the photolytic degradation of BaP. On the basis of their observations, they formed a theory that represents a combination of the three hypotheses listed by us. They suggested that, as undissociated BaP is very reactive, it readily enters into equilibrium involving complex formation upon activation by light or heat. The complex thus formed is possiblya dimer which locks the C-H out-of-plane vibrations to one side of the plane. This complex may enter into further complex formation through hydrogen bonding to hetero atoms . As BaP readily undergoes oxidation, this hydrogen bonding may occur also to oxidation products of the parent compound (15). Studies of the reactions between PAH adsorbed on solid surfaces and oxides of nitrogen and sulfur have been reported by several authors (9,14).
Even if our results are partly inconclusive with respect to what compounds are formed on the alumina, it seems fairly certain that most of the PAH present on secondary alumina is not given off to the ambient air upon heating.
A limitation of the experimental system is that the BaP spiked on the secondary alumina does not necessarily behave exactly as the PAH adsorbed during the cleaning process. However, if there is a difference, it would be most likely that the spiked P AH are less strongly adsorbed and consequently even more unlikely to evaporate to the work environment during the smelting process .
The validity of our conclusion may also be limited by the fact that the experimental system is different from the situation occurring when secondary alumina