Characterization of individual aerosol particles in workroom air of aluminium smelter potrooms

Abstract

Aerosol particles with aerodynamic diameters between 0.18 and 10 μm were collected in the workroom air of two aluminium smelter potrooms with different production processes (Søderberg and Prebake processes). Size, morphology and chemical composition of more than 2000 individual particles were determined by high resolution scanning electron microscopy and energy-dispersive X-ray microanalysis. Based on chemical composition and morphology, particles were classified into different groups. Particle groups with a relative abundance above 1% (by number) include aluminium oxides, cryolite, aluminium oxides–cryolite mixtures, soot, silicates and sea salt. In both production halls, mixtures of aluminium oxides and cryolite are the dominant particle group. Many particles have fluoride-containing surface coatings or show agglomerations of nanometer-sized fluoride-containing particles on their surface. The phase composition of approximately 100 particles was studied by transmission electron microscopy. According to selected area electron diffraction, sodium β-alumina (NaAl₁₁O₁₇) is the dominant aluminium oxide and cryolite (Na₃AlF₆) the only sodium aluminium fluoride present. Implications of our findings for assessment of adverse health effects are discussed.

---

Introduction

Aluminium (Al) is produced by electrolysis of alumina (Al₂O₃) to which cryolite (Na₃AlF₆) is added to lower the melting point. There are two technological processes in use in the Norwegian primary Al industry. In the Søderberg process, self-baking electrodes (called Søderberg electrodes) are used which graphitize under the heat of the electrolytic process. A major drawback of this method is that the electrolytic cells are not capsuled. Hence, heat is lost to a large extent and the air in the potrooms is considerably polluted by vapour and dust. In the Prebake process, the electrodes are graphitized prior to their use and the electrolytic cells operate under locks. Therefore, process heat, vapour and dust are greatly retained. The emission of polynuclear aromatic hydrocarbons (PAHs) is also largely reduced from pots operated by the Prebake technology.¹ A critical technological comparison of the Søderberg and Prebake process was provided recently.²

In both Al production technologies, airborne particles of variable composition (e.g., aluminium oxides, cryolite, soot), a number of gases (e.g., carbon oxides, sulfur dioxide (SO₂), hydrogen fluoride (HF)) and PAHs are generated. This mixture of contaminants is inhaled by workers in the reduction halls of an Al plant and may cause bladder cancer^3,4 and respiratory disorders.^5–7 Bronchial hyper-responsiveness (potroom asthma) is still occurring among workers in the primary Al smelters, but the pathophysiological mechanisms and the causative agent(s) involved are presently unknown.^8–10

The workroom air quality in the potrooms has traditionally been monitored by measuring a worker’s exposure to “total” dust and gases including particulate fluoride, HF, SO₂ and coal tar pitch volatiles (e.g., PAHs).⁸ Workers’ exposures in aluminium potrooms are not fully characterized. Most previous studies in the aluminium industry have either not collected the appropriate health related aerosol fractions (inhalable, thoracic, respirable) or have not investigated the chemical and phase composition of working place aerosols by microanalytical techniques. Bulk analytical methodologies have been applied to assess the exposure to total mass concentrations and especially for fluorides. Important parameters, for example, the phase composition or the mixing properties of the respirable and thoracic aerosol particles which penetrate the lung, are not known. In contrast, characterization of workroom aerosols by the combination of scanning and transmission electron microscopy (SEM and TEM) yields additional information including size, morphology, chemical and phase composition of individual particles. This combination of electron beam techniques has recently been applied successfully to characterize workroom aerosols in a nickel (Ni) refinery.¹¹

The present study was aimed at characterizing the chemical and morphological composition of respirable and thoracic particles present in workroom air of a primary Al smelter.

Experimental

Sampling took place in 1999 in two potrooms with different production processes (Søderberg and Prebake processes) of a primary smelter located at the west coast of Norway. The aerosol samples were collected on Ni TEM grids (with and without conducting Formvar or self-manufactured carbon foil) using a five-stage cascade impactor¹² in workers breathing zones next to the pots in both potrooms. The 50% cut-off diameters (aerodynamic diameter) of the five impactor stages are 0.18–0.35 μm, 0.35–0.65 μm, 0.65–1.2 μm, 1.2–3.5 μm, and 3.5–10 μm, respectively. At a flow rate of 2 l min⁻¹, the short-term sampling time varied between 30 and 1800 seconds (in order to obtain an optimum loading for electron microscopy). The pump employed was an in-house, pulsation-free unit constructed at the National Institute of Occupational Health (Oslo, Norway).

To minimize alteration, samples were initially not coated prior to scanning electron microscopy. In order to achieve a better quality of the secondary electron images for publication, typical images of the different particle groups were taken from two samples coated with a gold/palladium film after the first SEM inspection. All samples studied by TEM were carbon coated in order to avoid charging.

Size, morphology and chemical composition of the aerosol particles were investigated by high resolution SEM and energy-dispersive X-ray microanalysis. More than 2000 individual particles were studied in a field emission gun SEM (PHILIPS XL 30 FEG; Philips, Eindhoven, The Netherlands) and an environmental SEM (PHILIPS XL 30 ESEM LaB₆; Philips, Eindhoven, The Netherlands). Both instruments permit imaging of individual aerosol particles with geometric diameters above approximately 10 nm. X-Ray microanalysis of elements with Z ≥ 5 (boron) was carried out in both instruments with energy-dispersive Si(Li) detectors (TM DX4 and PHOENIX, EDAX, Tilburg, The Netherlands). Reliable energy-dispersive X-ray measurements were restricted in both SEMs to particles with diameters above 100 nm. The chemical compositions of the particles were only determined qualitatively, i.e., X-ray count rates were not corrected for matrix and geometric effects. Based on the qualitative chemical composition, the particles could be assigned to different groups (see below).

TEM was performed with two different instruments: a Philips CM12 (Philips, Eindhoven, The Netherlands) operated at 120 kV accelerating voltage and a Philips CM 20 ST operated at 200 kV. Both instruments were equipped with energy-dispersive X-ray detectors, which allow detection of elements with Z ≥ 6 (carbon). The CM12 was equipped with a Si(Li) detector (PV9900, EDAX, Tilburg, The Netherlands), and the CM20 with a high purity Ge detector (Thermo NORAN, Middleton, United States). Energy-dispersive X-ray microanalysis was restricted in both TEMs to particles with diameters above 50 nm. Phase identification is based on electron diffraction patterns which were indexed using the ICSD databank (February 2003, http://www.fiz-informationsdienste.de/en/DB/icsd/) and the software package PIEP.¹³

Results

The size distributions (equivalent projected area diameter) of particles in the workroom air from both potrooms are presented in Fig. 1. The low particle numbers for diameters below 300 nm (Fig. 1) are most likely an artifact of the sampling and analysis procedures. In a previous paper,¹⁴ we have shown by comparison with a scanning mobility particle sizer (SMPS) that our electron microscopical approach underestimates particle concentrations substantially in the size range below 300 nm. This can be explained by the inhomogeneous loading of the sampling substrate which leads to deposition of small particles outside the visible deposition spot and a low deposition efficiency for particles with diameters below 200 nm. However, in the Søderberg potroom a higher abundance of large particles (diameters ≥ 500 nm) is observed compared to the Prebake potroom (Fig. 1).

Fig. 1 Size distribution (equivalent projected area diameter) of particles in workroom air: (a) Søderberg hall (b) Prebake hall. The low particle numbers at diameters below 300 nm are most likely an artifact of the sampling and analysis procedures (see text for details).

Based on the qualitative chemical composition and morphology, the particles were assigned to different groups. Confidence intervals for the relative abundances of the different particle groups were calculated assuming a multinomial distribution:¹⁵

with v_(i) the confidence interval of particle group i, r_(i) the absolute number of particles in group i, n the total number of particles investigated in a sample, and χ²_k−1;1−α the critical value of the Chi-square distribution (k = number of particle groups, k − 1 = degrees of freedom, 1 − α = significance level).

The relative abundances and the 95% confidence intervals of the different particle groups are given in Table 1. Secondary electron images and energy-dispersive X-ray spectra of typical particles from the most abundant groups are displayed in Fig. 2. In both potrooms particles containing Al, sodium (Na), fluorine (F), and oxygen (O) are dominant with relative abundances of approximately 45% (Søderberg) and 65% (Prebake), respectively. This dominant particle group is most likely a mixture of aluminium oxide and cryolite (Fig. 2a and 2b) and can easily be recognized in the energy-dispersive X-ray spectra (Fig. 2b), as the above-mentioned elements are present as major components. The aluminium oxide–cryolite mixtures may also consist of small primary particles with diameters in the range of 0.1 and 1 μm which often form complex agglomerates with sizes up to a few μm. Many particles (see Table 1) predominantly consist either of aluminium oxide (Fig. 2c and 2d) or cryolite (Fig. 2e and 2f). The former can be recognized in the energy-dispersive X-ray spectra from the high Al and O peaks and the low abundance of Na and F (Fig. 2d), the latter by their low O contents (Fig. 2f). Many oxide grains occur as platelets with almost hexagonal morphology (Fig. 2g). On the surface of most aluminium oxide grains, smaller cryolite and/or aluminium oxide particles are aggregated (Fig. 2a, 2c, 2e and 2g). In this case, the presence of cryolite can be recognized in the X-ray spectra by the presence of a small F peak. These surface deposits/coatings are most likely caused by vapour phase formation of ultrafine particles. In a few cases, cryolite (monoclinic) occurs as needles. Soot particles are frequently encountered in both potrooms (Table 1). Soot agglomerates can easily be recognized from their typical morphology (Fig. 2i and 2k) described for combustion aerosols.^16–20 The agglomerates often exhibit fractal-like geometries. However, no attempt was made to determine their fractal properties (fractal dimension and fractal pre-factor). As can be seen from their complex chemical composition (Fig. 2l), the soot agglomerates may also contain considerable amounts of additional phases as for example aluminium oxides, cryolite, silicates, sulfides/sulfates. In addition to soot agglomerates, some large carbon-rich particles (up to a few μm) are observed which are most likely splinters of the carbon electrodes. As only a few grains were observed, these carbon-rich particles were assigned to the group of other particles. Silicates and sea salt are observed in both potrooms with a relative abundance between 1% and 6%. Brass particles (about 1.8% of all analyzed particles) are most likely an artifact of the sampling procedure since the sample holders consisted of this material. All other particle groups listed in Table 1 occur at abundances below 1% and are, thus, not considered any further. It is interesting to note that needle-like shapes were observed in all particle groups containing Al-phases and in the group of silicates. In the Søderberg and Prebake potrooms approximately 1.4% and 0.9% of all particles show needle-like shapes, respectively.

Fig. 2 Secondary electron images and energy-dispersive X-ray spectra of typical particles from the most abundant groups: mixtures of aluminium oxides and cryolite (a, b), aluminium oxide (c, d), cryolite (e, f), aluminium oxide platelets (g, h), pure soot agglomerates (i, j), and soot agglomerates with inclusions of other phases (k, l).

Table 1 Relative abundances [%] and 95% confidence intervals (in parenthesis) of particle groups

Particle group	Søderberg (1012 particles)	Prebake (1006 particles)
a Various phases, predominantly sodium β-alumina (see text for details).
Aluminium oxides^a	15.2 (10.9–20.9)	8.0 (5.0–12.4)
Cryolite	9.2 (5.9–14.0)	11.4 (7.8–16.4)
Aluminium oxides^a–cryolite mixtures	45.1 (38.3–52.0)	64.6 (58.0–70.8)
Soot	13.4 (9.4–18.9)	6.6 (3.9–10.7)
Silicates	6.1 (3.6–10.4)	1.3 (0.4–3.9)
Sea salt	5.1 (2.8–9.2)	4.5 (2.4–8.2)
Brass (sample holder)	1.7 (0.6–4.6)	1.8 (0.7–4.6)
Fe/Ti-oxides	0.6 (0.1–3.0)	0.8 (0.2–3.2)
Calcium carbonate	0.9 (0.2–3.4)	0.1 (0.005–2.0)
Sodium sulfates	0.9 (0.2–3.4)	—
Calcium sulfates	0.2 (0.02–2.3)	—
Silicon	—	0.1 (0.005–2.0)
Other particles	1.6 (0.6–4.5)	0.9 (0.2–3.3)

---

The phase composition of the Al-containing particles was studied by TEM. Since we have studied only about 100 particles by selected area electron diffraction (SAED), the abundance of the different phases is not known precisely and certainly needs further study. Most surprising is the fact that we did not observe any corundum (α-Al₂O₃). The only oxide phase identified unequivocally by SAED is NaAl₁₁O₁₇ (sodium β-alumina²¹); all particles which occurred as almost hexagonal platelets were found to consist of this phase. One example of a TEM bright field image and two electron diffraction patterns ([001], [011]] are shown in Fig. 3. As the predominant occurrence of sodium β-alumina is quite astonishing, we have studied the electron diffraction patterns in some detail. Diffraction patterns were obtained in six different orientations of the same grain and were simultaneously compared to a list (ICSD databank, February 2003) of all possible phases containing at least two of the elements Al, O, Na and F. They were indexed as [001], [1̄2̄1], [201̄], [011], [121], [1̄21] of sodium β-alumina (space group: P6₃/mmc; lattice constants: a = b = 5.594 Å, c = 22.53 Å). In addition, an orientation matrix was obtained and used to check the angles between the zone axes.

Fig. 3 TEM bright field image (a) and two electron diffraction patterns (b, c) of sodium β-alumina (NaAl₁₁O₁₇).

A few grains of pure aluminium oxide were encountered in our TEM work. According to SAED, these grains consist of AlO (comparison to the ICSD databank). However, it should be emphasized here that unequivocal differentiation between the various aluminium oxide phases was not always possible.

In addition to aluminium oxide phases, cryolite (Na₃AlF₆) was identified by selected area electron diffraction both as particles and needles. Small oxygen peaks occurring in the energy-dispersive X-ray spectra of these grains were interpreted as the result of the presence of small oxide grains on the surface of cryolite. One example of a TEM bright field image and an electron diffraction pattern of cryolite is given in Fig. 4.

Fig. 4 TEM bright field image (a) and electron diffraction pattern (b) of cryolite (Na₃AlF₆).

Discussion

An important finding of our study is that the workers in Al potrooms are exposed to complex agglomerates and particles consisting of aluminium oxides (with sodium β-alumina (NaAl₁₁O₁₇) as the dominant phase) and cryolite (Table 1). In the most abundant particle group (aluminium oxides–cryolite mixtures), both phases are present as major components. The less intense signals of Na and F from particles where the X-ray spectrum is dominated by Al and O, clearly indicate that these particles consist of a core of aluminium oxides and a relatively thin nanostructured surface coating of sodium fluorides. The chemical and phase compositions of these ultrafine particles (on the surface of larger particles) are not known, but theoretical considerations on formation of ultrafine particles in Al-potrooms²² suggest that these are mainly condensation and hydrolysis products from vaporization of the cryolite containing electrolyte. The dominant vapour phase species is NaAlF₄ which dissociates into solid chiolite (Na₅Al₃F₁₄) and AlF₃ on cooling.

Most surprisingly not only cryolite needles were found in our study, but also mixed cryolite–alumina ones. L’vov et al.²² suggest that the cryolite vapour may hydrolize to form Al₂O₃, NaF and HF. Alumina vapour in super-saturated conditions may form needle-like crystals. Thus, it is interesting to note that Voisin et al.²³ have identified high concentrations of “Al-fibres” in both lung tissue and bronchoalveolar lavage fluid from Al-potroom workers. They concluded that the most probable fibre composition was Al₂O₃ due to the absence of Na in their X-ray spectra. In contrast to Gylseth et al.²⁴ who reported the occurrence of fibrous NaAlF₄ in the workroom atmosphere of Al-potroom crust breakers, we have not been able to identify fibres with this composition. However, it should be noted that Gylseth et al. do not exclude the presence of other phases like cryolite and chiolite (instead of NaAlF₄) in their study.²⁴ These fibres showed no signs of chemical attack after 5 minutes sonication in either distilled water or 1 M hydrochloric acid.²⁴ This may suggest that even among those fibres, insoluble aluminium oxides could occur frequently.

Pure aluminium oxides and pure cryolite are almost absent in the samples under study. It is also surprising that we did not observe any corundum, especially since it is the major constituent of the alumina raw material. Since alumina used in the Al-industry contains minor or trace amounts (approximately 0.5%) of sodium β-alumina (NaAl₁₁O₁₇), it may be likely that the “fines” of the alumina raw material are enriched in this phase.²⁵ This fine fraction of alumina may thus be part of the thoracic/respirable aerosol fraction of the workroom air in potrooms. However, as the number of aluminium-containing particles studied by SAED was limited, further studies are needed in order to obtain better statistics for the abundances of the different phases. Most aluminium oxide agglomerates also contain some cryolite and vice versa. Thus, it can be concluded that exposure to single aluminium compounds does not occur.

The carbon-rich particles identified are mostly soot agglomerates that may originate either from the production process or from diesel-fueled vehicles in the potrooms. As the soot agglomerates are a major component of the aerosols collected in both potrooms and often contain inclusions of additional phases, this particle group cannot be neglected as a potential carrier of other potroom pollutants into the lung.

Our study clearly shows that the agglomerates identified often consist of small fluoride-containing primary particles with diameters in the range of 0.1 to 1 μm. Due to the small sizes of the primary particles, the agglomerates have a larger surface area that leads to an enhanced reactivity compared to μm grains of the same composition. The most important finding is undoubtedly the frequently observable fluoride-containing coating on aluminium oxide and cryolite particles. This coating exhibits a nanostructure with an enormous specific surface which most likely enhances the solubility in the lung. Furthermore, the agglomerates and the coated particles, to which SO₂ and HF can absorb, may very well be excellent vehicles for transporting these reactive gases into the thoracic region of the lung. The nano-layer of fluorine-containing deposits on most particles may also form HF after deposition in the lung or exposure to the respiratory humidified air.

Acknowledgements

Financial support is gratefully acknowledged from the Work Environment Fund of the Norwegian Business Association.

References

1. J. Damiano, in Health in the Aluminium Industry, ed. N. D. Priest and T. V. O’Donell, Middlesex University Press, London, England, 1998, pp. 8–20.
2. G. Nechaev, S. Tsymbalov, L. E. Swartling and G. E. Volfson, in An economic Söderberg to Prebake conversion for the Russian aluminium industry, in Light metals 1997, ed. R. Huglen, Proceedings TMS Annual Meeting, Orlando, FL, USA, 1997, pp. 201–205.
3. G. W. Gibbs, Ann. Occup. Hyg., 1997, 41, 49.
4. B. Armstrong, C. Tremblay, D. Baris and G. Theriault, Am. J. Epidemiol., 1994, 139, 250.
5. J. Kongerud, J. Boe, V. Søyseth, A. Naalsund and P. Magnus, Eur. Respir. J., 1994, 7(1), 165.
6. J. G. Gilson, A review of non-malignant respiratory disorders in primary aluminium smelters., International Primary Aluminium Institute, London, England, 1987.
7. M. L. Abramson, J. H. Wlodarizyk, N. A. Saunders and J. Hensley, Am. Rev. Respir. Dis., 1989, 139, 1042.
8. G. Benke, M. Abramson and M. Sim, Ann. Occup. Hyg., 1998, 42(3), 173.
9. B. Sorgdrager, A. J. A. de Looff, T. M. Pal, F. J. H. van Dijk and J. G. R. de Monchy, Int. Arch. Occup. Environ. Health, 2001, 74(1), 55.
10. K. Radon, D. Nowak, R. Heinrich-Ramm and D. Szadkowski, Int. Arch. Occup. Environ. Health, 1999, 72(5), 297.
11. S. Weinbruch, P. van Aken, M. Ebert, Y. Thomassen, A. Skogstad, V. P. Chashchin and A. Nikonov, J. Environ. Monit., 2002, 4, 344.
12. P. Wieser and R. Wurster, Application of laser-microprobe mass analysis to particle collections, in Physical and chemical characterization of individual airborne particles, ed. K. R. Spurny, Ellis Horwood Ltd., Chichester, UK, 1986, pp. 251–270.
13. G. Miehe, Ber. Dtsch Mineralogischen Ges. Beih. z. Eur. J. Mineral, 1997, 9, 250.
14. M. Ebert, S. Weinbruch, A. Rausch, G. Gorzawski, P. Hoffmann, H. Wex and G. Helas, J. Geophys. Res., 2002, 107 , NO. D21, 8121, doi:10.1029/2000JD000195.
15. J. Hartung, Statistik, 8th edn., Oldenbourg Verlag, München, Germany, 1991.
16. Ü. Ö. Köylü, G. M. Faeth, T. L. Farias and M. G. Carvalho, Combust. Flame, 1995, 100, 621.
17. Ü. Ö. Köylü, Y. Xing and D. E. Rosner, Langmuir, 1995, 11, 4848.
18. I. Colbeck, B. Atkinson and Y. Johar, J. Aerosol Sci., 1997, 28, 715.
19. A. L. Dye, M. M. Rhead and C. J. Trier, Atmos. Environ., 2000, 34, 3139.
20. M. Wentzel, H. Gorzawski, K.-H. Naumann, H. Saathoff and S. Weinbruch, J. Aerosol Sci., 2003, 34, 1347.
21. K. Edstroem, J. O. Thomas and G. C. Farrington, Acta Crystallogr., Sect. B, 1991, 47, 635.
22. B. V. L’vov, L. K. Polzik, S. Weinbruch, D. Ellingsen and Y. Thomassen, J. Environ. Monit. 10.1039/b501302j.
23. C. Voisin, F. Fisekci, B. Buclez, A. Didier, B. Couste, F. Bastien, P. Brochard and J.-C. Pairon, Eur. Respir. J., 1996, 9, 1874.
24. B. Gylseth, O. Bjørseth, Ø. Dugstad and J. Gjønnes, Scand. J. Work Environ. Health, 1984, 10, 189.
25. H. Kvande, Norsk Hydro Aluminium, Oslo, Norway, Personal communication, 2004.

---