Airborne Fibres in the Norwegian Silicon Carbide Industry

Transcription

1 Ann. Occup. Hyg., Vol. 50, No. 3, pp , 2006 Ó The Author Published by Oxford University Press on behalf of the British Occupational Hygiene Society All rights reserved. doi: /annhyg/mei081 Airborne Fibres in the Norwegian Silicon Carbide Industry A. SKOGSTAD 1 *, S. FØRELAND 1,2, E. BYE 1 and W. EDUARD 1 1 Department of Occupational Hygiene, National Institute of Occupational Health, PO Box 8149 Dep, N-0033 Oslo, Norway and 2 Department of Chemistry, University of Oslo, PO Box 1033, Blindern, N-0315 Oslo, Norway Received 29 June 2005; in final form 11 October 2005; published online 23 February 2006 Morphology of silicon carbide (SiC) fibres from the Norwegian SiC industry has been studied by scanning electron microscopy (SEM). The fibres are an unwanted side-product in SiC production. They represent a probable cause of the observed increased occurrence of lung diseases among SiC workers. The main aim of this work is to give a detailed description of the morphological variation of the fibres. Furthermore, it is important to study the occurrence of various morphological types with respect to job types and process parameters. SiC fibres accounted for >90% of all fibres observed. Eight categories of SiC fibres are described based on their morphology. The most frequent fibre category had a smooth surface and accounted for more than half of the observed SiC fibres. The diameter distributions of the eight fibre types were significantly different except for two of the categories. More than 99% of the SiC fibres observed were <3 mm in diameter, satisfying one WHO criterion for health relevant fibres. The aspect ratio and diameter of health-relevant fibres generally followed a lognormal distribution for different fibre categories, whereas fibre length did not. The proportions of SiC fibres (all categories) did not differ significantly between the plants. The proportions differed between plants for two SiC fibre categories including the most dominant type. For two SiC fibre categories and the SiC cleavage fragments differences were observed between job groups. Two other fibre categories were correlated with type of SiC produced (i.e. black or green SiC) and sawdust added to the raw material mix. Keywords: fibre dimensions; fibre morphology; job groups; scanning electron microscopy; silicon carbide fibres INTRODUCTION Silicon carbide (SiC) is a very hard material that is widely used as an abrasive in cutting tools and as raw material in the refractory, foundry and ceramic industry. SiC is produced from quartz and petrol coke in open electric resistance furnaces at a temperature of 2500 C. The production of SiC generates many airborne contaminants including crystalline and amorphous silica, SiC fibres, non-fibrous SiC particles, polyaromatic hydrocarbons, carbon monoxide (CO) and sulphur dioxide (Smith et al., 1984; Bye et al., 1985; Dufresne et al., 1987; Scansetti et al., 1992; Petry et al., 1994; Gunnæs et al., 2005). Hence, workers in the SiC industry may be exposed to a variety of chemical agents, *Author to whom correspondence should be addressed. asbjorn.skogstad@stami.no many of them representing significant risks to health. Diseases of the respiratory system are well documented among workers exposed to silica (Bruusgaard, 1948; IARC, 1987, 1997) and sulphur dioxide (Smith et al., 1977; Peters et al., 1984; Osterman et al., 1989). Increased risks of various cancer types have been observed in the SiC industry (Infante-Rivard et al., 1994; Romundstad et al., 2001). The exposure to crystalline silica as well as SiC fibres are possible causes for the increased risk of lung cancer, because both agents are recognized as human carcinogens (IARC, 1987, 1997; ACGIH, 2003; Norwegian Labour Inspectorate, 2003). Romundstad et al. (2001) found a dose response relationship between the standardized incidence ratio of lung cancer and the cumulative exposure to total dust. However, identification of any possible specific causative agents in the airborne dust for lung cancer was not achieved (Romundstad et al., 2001). 231

2 232 A. Skogstad et al. The assessment of SiC fibre exposure has until now mainly been carried out without speciation of fibre types, applying phase contrast optical microscopy and the WHO-method (WHO, 1997) for counting. It was evident from earlier studies that several types of fibres were present, with different chemical compositions and morphology. (Bye et al., 1985; Dufresne et al., 1987). Therefore, as part of an ongoing retrospective study of the exposure in the Norwegian SiC industry, a better characterization of airborne fibres in the work environment has been carried out. Recently, we have reported on the morphology and structure of SiC fibres in the SiC industry (Gunnæs et al., 2005). In that study the emphasis was on the branching phenomena of this fibrous material, with fibres grown together as seen in Fig. 1n. Here we report on the occurrence of fibre types in the work Fig. 1. SEM images of different SiC fibre morphologies (categories). (a c) K1 with variation of disc shape. (d) K2 with variable diameter along the fibre axis. (e) K3 fibre. (f) K4 fibre with smooth surface. (g) K4: pointed fibre end. (h) K4: fibre protruding from club-like structure. (i) K5 with irregular surface. (j) K6 fibre: the hexagonal cross section gives the impression of white lines along the fibre border. (k) Category K7. (l) Specimen of K8 fibre category. (m) Two cleavage fragments and one angular SiC particle. (n) Branched SiC fibre structure consisting of four branches of type K2. Scale bars represent 1 mm.

3 Airborne fibres in the Norwegian silicon carbide industry 233 environment classified by shape, chemical composition, and their length and diameter distribution. Possible determinants of the relative occurrence of the various fibre types, as plant, SiC type and raw material were also studied. the refinery department. The fibres are primarily found in the partly reacted material that is not transported to the refinery. Hence, we have collected personal fibre samples in the mixing, furnace and sorting departments. MATERIALS AND METHODS The SiC fibres studied in the present investigation were collected from all three SiC plants in Norway, i.e. Plant A, B and C. Green SiC (>99% SiC) is produced from a mixture of quartz sand and petrol coke. Black SiC (98% SiC) can be produced by use of reclaimed material and aluminium oxide in the mixture. Sawdust is added to the raw materials in Plant A to improve the porosity of the furnace mix. The mix is transported from the mixing department to the furnace department and loaded in an Acheson electrical resistance furnace, with a graphite core in the centre. The electrical heating raises the temperature of the graphite core to 2500 C. Silica will undergo a carbothermal reduction with petrol coke to produce SiC and CO in a gas phase reaction: SiO 2 þ 3C! SiC þ 2CO Electric power is supplied for 45 h, and the whole furnace cycle takes 8 10 days to complete. Unreacted material is then removed and returned to the mixing department while the SiC crude is transported to the sorting department. Here partly reacted material is removed from first grade crude and used in the metallurgic industry or recycled to the furnace mix. The first grade crude is crushed and transported to Job groups and job types Table 1 gives a summary of the different work activities in the SiC production industry. The SiC production in the three plants is based on the same technology; however, there are differences in the process and the way in which jobs are organized. A classification into three job groups can be done (see Table 1). Job group 1 includes mixing operator and pay loader who handle raw materials and reclaimed furnace mix. Job group 2 includes charger, furnace, crane, control room and mechanic operators who carry out furnace work. Job group 3 includes only sorting operators who sort the crude material. All three plants have job types included in Job group 2 and 3 while only plant B and C have Job group 1 (see Table 1). Air sampling Personal samples were collected in all the job types in the mixing, furnace and sorting departments using pulsation-free personal pumps constructed at NIOH (National Institute of Occupational Health, Oslo, Norway). Air sampling was performed within a 2 day period for each plant. A total of 32 air samples were collected, with 10 samples per plant (see Table 1). The sampling duration varied between 0.5 and 2 h. This sampling strategy was adopted for two reasons. Since polycarbonate filters are Table 1. Occupational job types, job groups and summary of different work activities in the Norwegian silicon carbide industry Job type Job group No. of samples Description of work Mixing operator (Plant C) Charger operator (Plant C) Furnace floor operator (Plant A) 1 3 The mixing operator is in charge of the mix building where the furnace mixture is made from quartz sand, coke and reclaimed furnace mix. He supervises the process from a control room, but regular inspections are necessary 2 2 The charger operator works both on the outside and inside of the furnace during loading 2 2 The furnace floor operator works on the outside and inside of the furnace during filling of raw material and removal of the furnace product Crane operator (Plant B) 2 2 The crane operator works in closed cabins supplied with fresh air, situated up under the roof of the furnace plant Payloader operator (Plant B and C) Control room operator (Plant A and B) Sorting operator (Plant A, B and C) 1 5 The payloader operator transports raw materials from storage rooms to elevators connected to the mixing building. He also drives inside the furnace building. The payloaders used are closed cabin vehicles 2 5 The control room operator controls the furnaces and the dust release to the environment with equipment situated in a control room. He performs supervision rounds in the furnace plant and works occasionally in cranes and on furnace plant floor 3 12 The sorting operator separates the furnace product so that the partially reacted material can be removed from the fully reacted SiC Mechanics (Plant C) 2 1 The mechanic is involved in all sorts of mechanical maintenance work in the furnace department

4 234 A. Skogstad et al. susceptible to particle loss, the over-loading of filters had to be avoided. The resulting filter samples did not show any loss of fibres. Some workers accomplish different job activities. These may last from <1 h to several hours during the day. Sampling was consequently restricted to periods with uniform job activities. Fibres for scanning electron microscopy (SEM) analysis were collected on 25 mm polycarbonate (PC) filters with pore size 0.4 mm using open-faced graphite-filled 25 mm filter holders with 50 mm extension tube (Gelman Air Monitoring Cassette, Gelman Sciences, Ann Arbor, MI, USA). The airflow was adjusted to 1 l min 1. Sample preparation The exposed PC filters were transferred to clean glass beakers. The inside of the corresponding cassette cowl was rinsed with 10 ml of 25% ethanol filtered through Millipore Millex Ò -GS, with 0.22 mm pore size. Then the liquid was collected in the same glass beaker. The filter with the rinsed liquid was sonicated for 3 min in an ultrasonic bath with a frequency of 35 khz and a high-frequency power of 225 W (Sonorex RK510H, Bandalin Electric, Berlin, Germany) to liberate the particles from the filter. Sub-volumes from the homogenized particle suspension were filtered using a funnel with a 15 mm internal diameter on PC filters. These had been coated with a 100 nm layer of platinum and had an effective pore size of 0.2 mm for SEM analysis. Filters were cut and specimens of size 6 6mm 2 were mounted on aluminium stubs using double-sided carbon adhesive discs. Additional spots of carbon cement were applied at each specimen corner to secure good conductivity to earth. All samples were coated with a nm platinum layer in a Balzers SCD 050 sputter coater (Balzers, Liechtenstein). A suspension of a reference material with SiC whiskers (Stanton et al., 1981) in 25% ethanol was filtered through 0.8 mm PC filter and coated with platinum. Samples were analysed applying SEM with the same conditions as the aerosol samples. SEM analysis A Jeol JSM-6400 (Akishima, Tokyo, Japan) connected to a Vantage DSI energy dispersive X-ray spectrometer (EDS)-system (Thermo NORAN, Middleton, WI, USA) equipped with a thin window detector (NORVAR) was used for the analysis. This allows us to detect elements with atomic number >5. Fibres were counted by the use of an accelerating voltage of 10 kv, a working distance of 15 mm, a magnification of 2000 and a spot size of 50 nm. In the slow scan mode we were able to detect fibres with a diameter of 0.07 mm. The same conditions on the SEM were used for EDS-analysis. Some fibres were re-analysed at 5 kv to control for carbon and oxygen peaks. An acceleration voltage of 20 kv was used to obtain an even better resolution, to study finer details. The counting fields, using a calibrated reticule directly on the SEM screen were selected at random throughout the exposed filter area. The outer 2 mm from the cutting lines were omitted. A screen magnification of or was used to measure fibre diameter. For fibres with a nonuniform cross section along the fibre axis, the diameter taken as the mean of the largest and smallest diameter was recorded. Fibre length was measured with a suitable magnification for the fibre in question. The SEM-analysis was performed according to the ASTM: D and VDI 3492 (VDI, 1991) protocols with only minor modifications. Statistical methods The Kolmogorov Smirnov test was used to analyse the distributions of the fibre dimensions. The nonparametric Kruskal Wallis test and subsequent Mann Whitney tests were used to compare the size distributions of the SiC fibre types. Analysis of variance (ANOVA) and multiple regressions were applied on arcsine transformed fibre proportion data (Snedecor and Cochran, 1989). ANOVA was performed with plant as grouping variable. Multiple regression analysis was applied with job types, production site and production parameters as independent variables. Only determinants with P-values <0.1 were included in the models. Furthermore, fibre types with proportions <1% were excluded from the calculations. RESULTS Morphology SiC fibres were classified in seven distinct categories according to their morphology as seen by SEM. Fibres that could not easily be assigned to any of these groups were pooled in a rest category. The various fibre categories are shown in Fig. 1 and described below. Fibres within these eight groups were counted and their frequencies are listed in Table 2. In addition to single SiC fibres, complex structures, i.e. branched fibres, were also observed, see Fig. 1n. In the mixing department a total of 160 fibres including branched structures were counted. In the furnace and sorting departments the numbers were 1314 and 962 fibres, respectively. Category 1 (K1, Fig. 1a c): These fibres were straight or bent and able to form branches. The fibres looked like a staple of discs, mostly perpendicular to the fibre axis, see Fig. 1a. The shape of the discs varied slightly and they were densely packed along the fibre axis, see Fig. 1b. Some fibres were only partially covered by discs, revealing a central core as seen in Fig. 1c.

5 Airborne fibres in the Norwegian silicon carbide industry 235 Table 2. Size parameters (in mm) for the different SiC fibres categories SiC fibre category K1 K2 K3 K4 K5 K6 K7 K8 CF All N AM D L AR Min D L AR Max D L AR Median D L AR GM D L AR GSD D L AR K S a D NS * NS *** NS NS n.a. n.a. NS *** L *** *** * *** * NS n.a. n.a. NS *** AR NS NS NS NS NS NS n.a. n.a. NS NS Fibres with at least one end obscured by other particles are excluded. Kolmogorov Smirnov test indicate distribution fitness of parameters. D = diameter, L = length, AR = aspect ratio (L/D), CF = cleavage fragments, n.a. = not applicable, N < 30. *P < 0.05, **P < 0.01, ***P < a Kolmogorov Smirnov test for deviation of lognormal distribution at significance level a=0.05. NS = not significant deviation from a lognormal distribution. Category 2 (K2, Fig. 1d): Fibres in this category were mostly straight with variable diameter along the fibre axis with repetitive wide and narrow regions. Some of the K2-fibres were very thin with a diameter at the visibility limit for SEM, i.e mm. Category 3 (K3, Fig. 1e): Straight, mostly smooth, fibres but with small spines or knobs scattered on the surface. Category 4 (K4, Fig. 1f h): More than half of the SiC fibres were classified into this category. These fibres were rectilinear but often tapered along its axis. They had a circular cross section and a smooth surface. The thickest fibre end was sometimes pointed (Fig. 1g). Some specimens had a club-like structure on one end, most often with a vermiculated surface ornamentation as shown in Fig. 1h. Fibres with such structures were excluded in the diameter and length distribution. The fibres of this category (K4) had diameters reaching the visibility limit for our SEM. Category 5 (K5, Fig. 1i): Specimens of this category resembled the fibres in Categories 3 and 4, but the surface was more irregular. Category 6 (K6, Fig. 1j): These fibres had a hexagonal cross section, without discs, but with indistinct, transverse steps. The diameter was uniform along the entire fibre length. Category 7 (K7): Figure 1k shows a typical fibre of this category. They had an angular cross-sectional shape with distinct steps along the fibre. Careful observation revealed faint transverse striations, probably representing twins or stacking faults in the fibre crystal. These fibres were seldom observed.

6 236 A. Skogstad et al. Category 8 (K8): Fibres not easily recognized as members of any of the other groups were placed in this category. Such a fibre is shown in Fig. 1l. Cleavage fragments (Fig. 1m): Non-isometric SiC particles satisfying the fibre counting criteria for length, diameter and aspect ratio (AR) were sorted into its own category. These fibrous structures probably originate from breakage of non-fibrous SiC crystals. Many cleavage fragments were too coarse to be counted as fibres. Branched SiC fibres (Fig. 1n): Branched fibres were observed in all samples. They consist of a varying number of fibre branches, connected at different angles. Their complex morphology can be explained by branching during crystal growth of the cubic structure of the SiC fibres (Gunnæs et al., 2005). A maximum number of nine fibre branches were observed. The present fibre counting protocols are not easily applied for these branched SiC fibres, e.g. the WHO-method. Fig. 1n shows a branched fibre, which consists of four fibre branches of type K2. Other fibre types Less than 10% of all the fibres counted were non- SiC fibres. Most frequent were carbon fibres, comprising >30% of all the fibres in one sample. The other fibre types had mean proportions of <1% and were identified by XRMA as silicon oxide fibres (SiO), silicon fibres (Si), vanadium-rich fibres (V) and man-made vitreous fibres (MMVF). Dimensions Figure 2 shows the length-weighted diameter distribution of fibres selected according to the WHO counting criteria (D < 3 mm, L > 5 mm, L/D > 3:1) for the six most frequent categories of the SiC fibres weighted against fibre length. Except for cleavage fragments practically all the SiC fibres (99.9%) were <3 mm in diameter. Fig. 3a c shows the normal probability plot for log-transformed diameter (D), length (L) and AR for all SiC fibres, respectively. Applying the Kolmogorov Smirnov (K S) test on the cumulative AR for the whole material showed a lognormal distribution (Table 2). The diameter and especially the length distributions did not exhibit such a linear pattern. As the SiC fibres consist of morphologically quite different sub-populations, this distribution pattern can be expected. When the a b c Fig. 2. Length-weighted diameter distribution of the most frequent categories of SiC fibres. Fig. 3. (a c) Normal probability plots of fibre diameter (D), length (L) and aspect ratio (AR); log-transformed values, for all SiC fibres (N = 2263).

7 Airborne fibres in the Norwegian silicon carbide industry 237 K S test was applied on the separate SiC categories, the AR followed lognormal distributions for all the sub-populations of the fibres, as did the diameters for K1, K3, K5, K6 and the cleavage fragments. The fibre length showed linear distributions only in two cases, namely for K6 and the cleavage fragments. The deviations are most probably explained by the truncation at 5 mm for length and the visibility limit of the fibre diameter at 0.07 mm. Especially for K2 and K4 containing the thinnest fibre diameter the visibility limit had substantial influence (Table 2). Truncation of fibre counts at a diameter of 3 mm had little effect on the diameter distribution. The diameter, length and Table 3. Proportions of SiC fibre categories and other fibre types in three SiC plants in Norway Fibre type Proportions (%) All (n = 32) A (n = 10) B (n = 10) C (n = 12) SiC K K K K * K *** K ** K K CF Branched C * SiO Si V MMVF ANOVA Figures indicate the back transformed means of arcsinetransformed proportions in samples per plant. *P < 0.05, **P < 0.01, ***P < 0.001, n = number of samples. AR distributions are further summarized in Table 2. A pilot study was performed to check for breakage of fibres during sample preparation. We compared the length of directly counted fibres with sonicated fibres using light microscopy. The length geometric mean decreased from 8.2 to 7.5 mm indicating some breakage of the fibres during preparation. This may have increased fibre counts, but fibres may also have been lost when their length became <5 mm. Owing to the deviations from lognormality nonparametric tests were used to further explore and compare the dimensions of the fibre categories. Except for the categories K2 and K4 (P = 0.4) the other categories revealed significant different diameter distributions according to Kruskal Wallis and subsequent Mann Whitney tests. The median of the diameter varied from 0.25 mm for the K2 and K4 fibres to 1.5 mm for the cleavage fragments. The median of the length for all SiC fibres was 8.2 mm, with small variation between the various categories. The length distribution of K2 fibres was significantly different from all the other categories except for the cleavage fragments. Further, the tests showed differences between the length distributions for K1 compared with K4 and cleavage fragments. The ARs were significantly different in four cases, namely K1 and K6, K2 and K4 and K7 compared with K3 and K4. The median of the AR varied from 10.5 for K6 to 33 for K4. Determinants for the fibre type occurrence The proportions of the various fibre categories within the three SiC plants were compared (Table 3). Category K5 (P < 0.001) and C-fibres (P < 0.05) revealed the largest inter-plant difference with a high proportion in Plant C. In contrast the categories K4 and K6 revealed the lowest proportions in Plant C. Multiple linear regressions were used to study relationships between the occurrences of fibre types. The production determinants plant, job group (work Table 4. Effects of determinants on the occurrence of fibre types in the silicon carbide industry in Norway Fibre type a n R 2 Adj. Proportions (%) Plant Job group SiC type Sawdust A B C Green Black Yes No K K K K K K CF Branched C-fibre a Only fibre types with proportions >1% are included.

8 238 A. Skogstad et al. Table 5. Proportion of branched fibres within the SiC fibre categories and plants SiC fibre category Number of branches Total number of fibres a type), SiC type (green or black) and the use of sawdust in the furnace mixture were analysed. In models of K4, K5 and branched fibres the determinant plant gave the best fit, explaining 22, 46 and 9% of the variance, respectively (Table 4). Job group gave the best model fit for SiC fibre categories K1, K2 and cleavage fragments. Job group explained almost 50% of the variability for the cleavage fragments. The highest proportion of cleavage fragments was found in the job group containing the job type sorting. K3 fibres were associated with the type of SiC produced, the highest proportions being found during production of black SiC. K6 fibres and carbon fibres were associated with sawdust (P < 0.05), with the lowest proportions of these fibres for the plant using this additive (Plant A). The occurrence of branched fibres among the plants and SiC fibre categories was tested by two-way ANOVA on the arcsinetransformed proportions. Both factors were significant (P < 0.05) but not the interaction term (P = 0.1). The unadjusted proportions are shown in Table 5. Branched fibres were most common in Plants A and C and within the fibre categories K1 and K3. DISCUSSION Proportions of branched fibres (%) Plant A B C K K K K K K K K a Cleavage fragments do not form branches and are not included. The present study showed that the airborne SiC fibres formed during the industrial production of SiC exist in several morphological categories. These can be distinguished by SEM. In addition, complex branched fibre structures were observed. Such complex structures can be explained by the cubic crystalline structure (Gunnæs et al., 2005) and were observed in all samples. One fibre category (K4) comprising more than half the observed fibres exhibited a smooth surface. The diameter distributions of the eight fibre categories were significantly different except for two of the fibre categories including the most frequent fibre type. These two categories also contained the thinnest fibres. Diameter and length distributions for the total SiC fibre material deviated significantly from the lognormal distribution whereas the AR did not. However, some separate categories followed lognormality for the diameter and length as well. These deviations might be explained by the truncation at 5 mm for length (due to criteria for fibre counting) and 0.07 mm for diameter (the visibility limit). This is important for the two categories with the thinnest fibres. Practically all fibres observed were <3 mm in diameter, which means a high probability for deposition in the lungs. SiC fibres comprised >90% of all fibres observed. Carbon, silicon oxide, silicon, vanadium-rich fibres and MMVF made up the rest. The latter fibre type originates from the insulation material used in the construction of the furnace. The proportions of SiC fibres differed between the plants for two categories, i.e. K4 and K5, and the branched SiC fibres. Multiple linear regression analysis showed that the occurrence of the most frequent SiC category (K4) varied among the plants. Differences between the job groups were found for three SiC fibre types, i.e. K1, K2 and the cleavage fragments. For cleavage fragments, job group explained 50% of the variability, with the highest proportion associated with the job type sorting. This was expected, as the mechanical crushing of the crude in this process is the most likely explanation for the formation of cleavage fragments. One category (K3) was associated with the type of SiC produced. The highest proportion of this category was found in samples taken during production of black SiC. Morphological variation of the SiC fibres as well as the presence of other silicon-containing fibres may be attributed to the complex production process. This process takes place over a long period. During this period the temperature in the furnace mix rises gradually to 2500 C and falls to room temperature afterwards. Construction of the furnace with a central graphite core results in a circular symmetric temperature gradient perpendicular to the long axis of the furnace. This causes a complex progress of chemical gas phase reactions and subsequent sublimation of the SiC produced. The following recrystallization gives the desired coarse a-sic. A zone of b-sic and partly reacted or unreacted raw material is present outside the layer of a-sic. Silicon-rich gas products may diffuse through this layer, with possible secondary reactions and condensation reactions. It is most likely that the SiC fibres originate in this outermost layer of the crude. This was verified by an accumulation of fibrous SiC in this layer, as observed with SEM and transmission electron microscopy (TEM) (Gunnæs et al., 2005). The recycling of partly reacted or unreacted raw material into the next furnace mixture when producing black SiC may also contribute to the morphological variation of SiC fibres.

9 Airborne fibres in the Norwegian silicon carbide industry 239 Fig. 4. (a c) Specimen from reference material of SiC whisker (Stanton et al., 1978, 1981), showing similarities with the Norwegian material (cf. Fig. 1). Scale bars represent 1 mm. Stanton et al. (1978, 1981) and Johnson et al. (1992) carried out in vivo and in vitro tests with various types of SiC fibres. These fibres revealed high biological activities, even higher than crocidolite. Miller et al. (1999) performed intraperitoneal tests on rats using a range of man-made mineral fibres and amosite asbestos. The SiC fibres used in their experiments containing 8% of branched fibres, produced mesotheliomas earlier and at a faster rate than the other fibre types tested. The morphology of the fibres used by Stanton et al. (1981) shows similarities to the Norwegian material studied in the present investigation. Fibres in a sample of the original Stanton material are similar to some categories in the present study (Fig. 4a c), despite completely different formation processes. Corresponding similarities exist between the diameter distributions, as Fig. 5. (a c) Diameter distribution of the SiC whiskers used by Johnson et al. (1982); (SiC #1, SiC #2 and SiC #3), Stanton et al. (1978, 1981) and the Norwegian material (NIOH). shown in Fig. 5. The fibre size distributions in the materials of Stanton and Johnson were determined by TEM that detects thinner fibres than SEM. After adjusting the fibre data to comparable dimension ranges (e.g. D > 0.1 mm, L > 5 mm) the proportions of so-called Stanton fibres were estimated. This fibre fraction (L > 8 mm and D < 0.25 mm) was associated with the highest tumour risk. The test material of Stanton et al. (1978, 1981) contained 35% of these fibres. The three samples of Johnson et al. (1992) contained 7, 5 and 16%, whereas we found 15% in the Norwegian SiC industry. This fully demonstrates the close resemblance of morphology and size distributions between the tested materials and the Norwegian airborne industrial samples. These findings support the cancer risk that the industrial SiC fibres represent and an occupational exposure limit set to 0.1 fibres per cm 3 (ACGIH, 2003; Norwegian Labour Inspectorate, 2003). CONCLUSION In the present fibre study, a range of distinct morphological types of SiC fibres were found in the SiC industry, all identified as SiC fibres. They made up >90% of all fibres found in the working atmosphere. The different morphologies were also manifested by different diameter distributions. Within the categories the fibre diameter and AR followed a lognormal distribution with few exceptions. The proportions of the different SiC fibre categories differed between plants, job groups and production parameters. The explanation for these conditions is still unclear. The morphologies as well as the proportions of Stanton fibres were comparable with the synthetically produced SiC fibres reported to be highly durable and carcinogenic in test animals (Stanton et al., 1978, 1981; Johnson et al., 1992; Miller et al., 1999). Furthermore, all SiC fibres counted were in the respirable range with high probability of deposition in

10 240 A. Skogstad et al. the lower airway region. The increased prevalence of lung cancer as reported by Romundstad et al. (2001) may, thus, be attributed to the SiC fibres. Acknowledgements We would like to thank Per Ole Huser and Kistian Kruse for technical assistance during the sampling of airborne dust at the work place. Norunn Torheim is greatly acknowledged for producing the photomontages. We highly appreciate the co-operation with the three Norwegian SiC production plants: K. S. Orkla Exolon and the two SiC plants in Lillesand and Eydehavn within Saint-Gobain Ceramic Materials. This project could not have been accomplished without the positive attitude, and valuable and necessary contribution from the workers in the three plants, carrying the sampling equipment. Finally we thank Dr L. E. Lipkin for sharing some of Dr M. F. Stanton s SiC fibre sample. We received the material in the 1980s. This material was especially made for Dr M. F. Stanton by General Technologies Corp., Rescon, VA, USA. The Norwegian Government and The confederation of Norwegian business and industry work environmental fund financially supported this investigation. REFERENCES ACGIH. (2003) Threshold limit values for chemical substances and physical agents & biological exposure indices. Cincinatti, OH: American Conference of Governmental Industrial Hygiene. ISBN ASTM. (1996) Standard test method for determining concentration of airborne single-crystal ceramic whiskers in the workplace environment by scanning electron microscopy. Report No. D USA: American Society for Testing and Materials. Bruusgaard A. (1948) Pneumoconiosis in silicon carbide workers. Proceedings of the 9th International Congress on Industrial Medicine. London, UK: Wright Briston. pp Bye E, Eduard W, Gjønnes J et al. (1985) Occurrence of airborne silicon carbide fibers during industrial production of silicon carbide. Scand J Work Environ Health; 11: Dufresne A, Perrault G, Sébastien P et al. (1987) Morphology and surface characterisation of particulates from silicon carbide industries. Am Ind Hyg Assoc J; 48: Dufresne A, Sébastien P, Perrault G et al. (1992) Pulmonary clearance of fibrous and angular SiC particulates in the sheep model of pneumoconiosis. Ann Occup Hyg; 36: Gunnæs AE, Skogstad A, Olsen A et al. (2005) Morphology and structure of airborne b-sic fibres produced during the industrial production of non-fibrous silicon carbide. J Mater Sci; 40: IARC. (1987) IARC Monographs on the evaluation of carcinogenic risks to humans: silica and some silicates, Volume 42. Lyon, France: International Agency for Research on Cancer (WHO). IARC. (1997) IARC Monographs on the evaluation of carcinogenic risks to humans: silica, some silicates, Coal Dust and para-aramid Fibrils, Volume 68. Lyon, France: International Agency for Research on Cancer (WHO). Infante-Rivard C, Dufresne A, Armstrong B et al. (1994) Cohort study of silicon carbide production workers. Am J Epidemiol; 140: Johnson NF, Hoover MD, Thomassen DG et al. (1992) In vitro activity of silicon carbide whiskers in comparison to other industrial fibers using four cell culture systems. Am J Ind Med; 21: Miller BG, Searl A, Davis JMG et al. (1999) Influence of fibre length, dissolution and biopersistence on the production of mesothelioma in the rat peritoneal cavity. Ann Occup Hyg; 43: Norwegian Labour Inspectorate. (2003) Occupational limit values (In Norwegian). Available from: arbeidstilsynet.no/regelverk/veiledninger/full361.html#19 Osterman JW, Greaves IA, Smith TJ et al. (1989) Work related decrement in pulmonary function in silicon carbide production worker. Br J Ind Med; 46: Peters JM, Smith TJ, Bernstein L et al. (1984) Pulmonary effects exposures in silicon carbide manufacturing. Br J Ind Med; 41: Petry TH, Schmid P, Schlatter CH. (1994) Exposure to polycyclic aromatic hydrocarbons (PAHs) in two different silicon carbide plants. Ann Occup Hyg; 38: Romundstad P, Andersen A, Haldorsen T. (2001) Cancer incidence among workers in the Norwegian silicon carbide industry. Am J Epidemiol; 153: Scansetti G, Piolatto G, Botta GC. (1992) Airborne fibrous and non-fibrous particles in a silicon carbide manufacturing plant. Ann Occup Hyg; 36: Smith TJ, Peter JM, Reading JC et al. (1977) Pulmonary impairment from chronic exposure to sulfur dioxide in a smelter. Am Rev Respir Dis; 116: Smith TJ, Hammond SK, Laidlaw F et al. (1984) Respiratory exposure associated with silicon carbide production: estimates of cumulative exposure for an epidemiological study. Br J Ind Med; 41: Snedecor GW, Cochran WG. (1989) Statistical methods. 8th edn. Ames, IA: Iowa State University Press. ISBN Stanton MF, Layard M. (1978) The carcinogenicity of fibrous material. In Proceedings of the workshop on asbestos: Definitions and measurement methods; held at the National Bureau of Standards, Gaithersburg, MD, Washington, DC: National Bureau of Standards, pp (NBS special publication no. 506). Stanton MF, Layard M, Tegeris A et al. (1981) Relation of particle dimension to carcinogenicity in amphibole asbestos and other fibrous minerals. J Natl Cancer Inst; 67: VDI. (1991) Measurement of Inorganic Fibrous Particles in Ambient Air. Scanning Electron Microscopy Method. VDI 3492 Part 1. WHO. (1997) Determination of airborne fibre number concentration. A recommended method by phase contrast optical microscopy (membrane filter method). Geneva: World Health Organisation. ISBN