PM_2.5 Emissions from Aluminum Smelters: Coefficients and Environmental Impact

ABSTRACT

From 2004 to 2009, aiming to better understand implications for its smelters, Rio Tinto Alcan conducted a detailed study of PM_2.5 and PM₁₀ (particulate matter [PM] ≤ 2.5 and 10 μm in aerodynamic diameter, respectively) in its facilities. This involved a two-level study: part 1, emission quantification; and part 2, assessment of aluminum smelter contribution to the surrounding environment. In the first part, U.S. Environmental Protection Agency Other Test Method (OTM) OTM27 and OTM28 are assessed as relevant and efficient methods for measuring fine particle emissions from aluminum smelter stacks. Rio Tinto Alcan has also developed a safe and robust method called CYCLEX to measure PM_2.5 and condensable particulate matter (CPM) at the roof vents of potrooms. This work aims to determine the PM_2.5 emission coefficients of 17, 55, and 417 g·t⁻¹ of aluminum produced (including CPM) in anode baking furnace exhaust (fume treatment center), at potroom scrubber stacks (gas treatment centers), and at potroom roof vents, respectively. Results indicate that roof vents are the primary PM_2.5 emitters (85% of all smelter emissions) and that 71% of all smelter PM_2.5 comes from CPM. In the second part, preliminary inorganic speciation studies are conducted by scanning electron microscopy–energy-dispersive X-ray analysis and by isotopic ratios to track smelter emissions to their surrounding environment. This paper releases the first speciation results for an aluminum smelter, and the preliminary isotopic ratio study indicates a 3% impact in terms of PM_2.5 emissions for a representative smelter in an urban area.

IMPLICATIONS

Aluminum smelters tend to continuously improve their competitiveness by incrementally increasing production. In this context, assessing the effect of major contaminants is overriding, and ambient air modeling is often the preferred way to do so. Fine particles fit this category, and the primary aluminum industry needs to accurately know their emission factors to obtain representative modeling. Moreover, not all aluminum smelters have a method to measure PM_2.5 at roof vents, the primary emission outlets. Therefore, this paper describes the first-rate PM_2.5 measurement methods for aluminum smelter roof vents without down-comers. It also provides insight for environmental managers for tracking PM_2.5 emissions in plant surroundings.

INTRODUCTION

Airborne particulate matter (PM) is composed of primary aerosols (emitted as such, including condensable particulate matter [CPM]) and secondary aerosols originating from reactions among gases (oxides of sulfur and nitrogen, volatile organic compounds [VOCs], and ammonia).1,2 There is now consensus in the international community that PM less than 2.5 μm in aerodynamic diameter (PM_2.5) can have adverse effects on human health and that, unfortunately, they are positively associated with hospital admissions.3–6 However, it is unclear which PM_2.5 components could be responsible for such toxicity.7 Nevertheless, the U.S. Environmental Protection Agency (EPA) and the European Union (EU) are implementing regulations aiming to reduce mortality-related to PM_2.5 exposure.8 Only a few studies currently focus on PM_2.5 emissions from heavy industries such as steel plants,9 foundries,10 and secondary aluminum smelters,11 but the primary aluminum industry has yet to be the subject of study.

The process of aluminum production involves the electrolytic reduction of alumina in a cell that is dissolved in a molten bath mainly containing cryolite. Thus, hundred-meter-long potrooms are subject to dust emissions associated with combustion gases. Raw pot gases and dust are recovered right from the cells and later efficiently scrubbed at the gas treatment center (GTC), most often with bag filters. During operations on pots (anode replacements in particular), PM that is neither recovered nor treated is emitted as the hoods are opened. These emissions escape through the roof vents. Significant dust is also emitted from anode baking furnaces, where prebake technology anodes are prepared. During this process, gas and dust are collected and treated at the fume treatment center (FTC), also with bag filters.

In the aluminum industry, total particulate matter (PM_total =PM>0.3 μm plus condensables) and usually PM₁₀ (PM ≤ 10 μm in aerodynamic diameter) are monitored in the ambient air around plants, and PM_total (but without condensables) is tracked at stacks and roof vents for reporting purposes. However, condensables are never taken into account, and the extent to which they are released is completely unknown. Furthermore, the measurement methods at stacks and roof vents in the aluminum industry are not up to date and should be reviewed in the light of scientific advancements. Many years ago EPA released stack methods for fine particle measurements; for example, filterable PM₁₀ and PM_2.5 are measured at stacks with EPA method Conditional Test Method (CTM) CTM04012, and CPM is estimated with EPA202.13 These methods are suitable for aluminum smelters with respect to their dry emissions and the sampling of unnecessary dilutions. EPA is now proposing two new methods for fine particle measurements: Other Test Method (OTM) OTM2714 and OTM28.15 OTM27 could be seen as the only official method for measuring PM₁₀ and PM_2.5 at stacks, superseding CTM040. It does not present any major changes. OTM28 is the new method for condensable measurements, in lieu of EPA202 and should not overestimate CPM as EPA202 does. Nevertheless, these methods are inapplicable for vented roofs of aluminum smelter pot-rooms. For smelters equipped with downcomers (collecting raw pot gases in potrooms—see Figure 1), method EPA14 can be adapted with CTM040 and EPA202 to measure PM₁₀, PM_2.5, and CPM at roof vents. For smelters without downcomers, Rio Tinto Alcan proposes a new method called CYCLEX.

Figure 1. Illustration of an electrolysis pot with the raw pot gases recovered and treated at the GTC with alumina. Unrecovered pot gases are vented through the roof of each potroom. This is where CYCLEX, which measures PM_2.5 and CPM, is located.

In anticipation of future PM_2.5 regulations and to better understand the potential environmental footprint of the aluminum industry, Rio Tinto Alcan launched its PM_2.5 program in 2004. For part 1 of the program, the objectives were (1) to identify the best PM_total and PM_2.5 measurement methods at stacks, and (2) to develop an innovative method called CYCLEX to quantify PM_total and PM_2.5 at roof vents. For part 2 of the program, the objectives were (1) to address the speciation of PM_2.5 emissions from a typical smelter, and (2) to track PM emissions from the smelter to the environment. To address these questions, some state-of-the-art technologies were used, such as scanning electron microscopy (SEM) and stable and radiogenic isotope analyses.

EXPERIMENTAL PROCEDURES

Note that, in this paper, the metric system is used for measurement units. Also, all statistical tests were performed using GraphPad Prism, version 4.00 for Windows (GraphPad Software; http://www.graphpad.com).

Aluminum Smelters Studied

The two smelters studied are Sebree Works (Kentucky, United States) and Alma (Quebec, Canada), both operating prebake technologies (P-155 and AP-30, respectively) and producing 190 and 420 kt of aluminum per year, respectively. Operations are extremely stable in time and thus any time window along the year is acceptable for sampling.

Part 1: Quantification of PM_2.5 Emitted by the Sebree Smelter

Part 1.1: Emissions from Stacks

To find out the PM_2.5 emission factors for the Sebree Works GTC (reduction potlines) and FTC (anode baking furnace), a local contractor conducted initial trials according to EPA methods CTM040/202 in June 2006 (four and five replicates, respectively, at GTC and FTC). A new sampling campaign was run in October 2008 using EPA OTM27 and OTM28 (three replicates). The aim of this second campaign was to compare the effectiveness of wet (EPA 202) and dry impingers (OTM28). Indeed, an artifact is known to occur with high concentrations of sulfur dioxide (SO₂),16 the latter being prominent at GTC stacks (100–150 parts per million [ppm]).

Part 1.2: Emissions from Roof Vents

Concerning monitoring at smelter roof vents, Rio Tinto Alcan designated Sebree Works as an ideal place to test an innovative way (CYCLEX) to measure PM_2.5. Compared with Canadian aluminum smelters, U.S. facilities are designed with sampling manifolds, or “downcomers,” at roof vents. These allow the smelters to measure PM emissions at roof vents with EPA M1417/CTM040/202 methods (hereinafter M14*); in other words, downcomers are gas collectors in specific potroom sections merging into one main duct and finally exhausting like a regular stack. Rio Tinto Alcan can then compare its CYCLEX candidate method to this reference method in potrooms. But with no downcomers, it is impossible to use the EPA M14* method because of the absence of a stack and the strictly forbidden presence of water in potrooms (risk of explosion).

The CYCLEX method basically involves inserting a curved sampler inlet from method 5A in the APEX PM_2.5 sizer (Figure 2). The PM_2.5 sizer may be attached to the end of a rod that can be swiveled from the catwalk. A pulley may be used where there are no cat-walks. A Teflon line then connects each modified APEX PM_2.5 sizer to a sampling train with wet impingers (EPA 202; dry impingers were not tested). With this design, PM₁₀ measurements are not possible. The setup involves one CYCLEX sampling train between two pots along an eight-pot section (four modified APEX PM_2.5 sizers). The four sampling trains are centered on a gas velocity anemometer and located just above the cat-walk. Figure 1 depicts the location of one CYCLEX arrangement; a potroom in Sebree Works has 64 pots. Results are thus expressed taking into account (1) the average airspeed inside of the potroom, measured with anemometers (at Sebree usually ∼1.6 m·sec⁻¹); and (2) the total open roof vent surface (at Sebree 4855 m²). CYCLEX tests were conducted side by side with the M14* method in 2007 (both 12 replicates) and 2008 (20 and 5 replicates, respectively). This candidate method is currently going through the approval process with Environment Canada and EPA Québec.

Figure 2. Picture of the modified PM_2.5 sizer for gas recovery in a potroom.

Part 2: Speciation of PM_2.5 in the Surrounding Environment of the Alma Smelter

The purpose of this part is to obtain first insights into the aluminum industry's PM_2.5 speciation. This study also aims to track PM_2.5 emissions from raw anodes through baking (FTC emissions) and anode consumption (GTC emissions) and lastly to the ambient air by using isotopic ratios. To do so, anodes were sampled because they are one of the main sources of fine particle emissions; anodes are made of carbon and pitch. Next, the FTC and GTC stacks were sampled in triplicate at 16.7 L/min (6- and 24-hr samplings, respectively, with Hi-Vol PM₁₀ retrofit to PM_2.5); the urban monitoring stations 1A (218,421 m east; 5,381,784 m north; elevation 100 m; 2000 m from the smelter) and 3A (215,053 m east; 5,384,914 m north; elevation 110 m; 1750 m from the smelter) were the last two sites for Hi-Vol PM_2.5 sampling (167 hr, including the FTC and GTC sampling periods, one replicate). These campaigns were conducted in May 2008; roof vents were not sampled because their emission signature is the same as that of the GTC. Prevailing winds and locations are shown in Figure 3. Each collected filter (FTC, GTC, 1A, and 3A) was analyzed in four ways. First, one quarter of the cellulosic filters was analyzed by SEM–energy-dispersive X-ray (SEMEDX) analysis under low vacuum and uncoated (Arvida Research & Development Centre [ARDC] metallographic laboratory). Second, another quarter of the filter was acid digested for inductively coupled plasma (ICP)–atomic emission spectroscopy (AES) analysis, and a third quarter was caustic digested for fluoride analysis by a continuous flow analyzer (ARDC analytical team).

Figure 3. Location of Alma smelter and its two air monitoring stations. Arrows indicate the prevailing winds.

Stable and radiogenic isotopes were analyzed on the last quarter of the cellulosic filter by the Université du Québec à Montréal (UQAM-GEOTOP) as follows: (1) reverse polarity thermo-ionization mass spectrometry (N-TIMS) for osmium,18 (2) multicollector–ICP–mass spec-trometry (MC-ICP-MS) for lead,19 (3) isotope ratio mass spectrometry (IRMS) for carbon and nitrogen,20 and (4) TIMS for strontium and neodymium.21,22

RESULTS AND DISCUSSION

Part 1: Quantification of PM_2.5 Emitted by the Sebree Smelter

One should note that all results are expressed in grams of PM by metric ton of aluminum produced. This ratio approach is mandatory to adequately compare PM emissions between different sites of aluminum production. In addition, only results falling in the 80–120% isokinetics are considered in this study.

Part 1.1: Emissions from Stacks

First, Table 1 clearly indicates that using method 202 or OTM28 for CPM measurements largely influences the results. Indeed, because CPM are included in PM_2.5 and PM_total, differences observed between the 2006 and 2008 results can only be explained by a 60% decrease in CPM results for GTC and FTC. This highlights the importance of an accurate CPM assessment, given that they make up 55 and 95% of GTC and FTC PM_total emissions, respectively (Table 2). Furthermore, total CPM are reduced in parallel with inorganic CPM content (data not shown), which suggests that inorganic gases are the main source of secondary aerosols for aluminum smelters, the industry being well-known to emit hydrogen fluoride (HF) and SO₂.23

Table 1. Comparisons of mean PM emission coefficients ± standard error of the mean at the Sebree smelter between (1) EPA CTM040/202 and OTM27/28 at GTC and FTC, and between (2) the M14* and Rio Tinto Alcan CYCLEX candidate method at roof vents

	CPM	PM2.5	PM10	PMtotal
FTC
2006 (n = 5) EPA CTM040/202	45.4 ± 7.9	47.0 ± 8.0	47.5 ± 8.1	47.8 ± 8.2
2008 (n = 3) OTM27/28	16.8 ± 4.3	17.1 ± 4.1	17.3 ± 4.1	17.6 ± 4.0
GTC
2006 (n = 4) EPA CTM040/202	121 ± 44	130 ± 46	137 ± 45	149 ± 43
2008 (n = 3) OTM27/28	48 ± 7	55 ± 7	63 ± 7	86 ± 12
Roof vents
2007 and 2008 (n = 17) M14*	273 ± 30	417 ± 39	—	1308 ± 129
2007 and 2008 (n = 32) CYCLEX	251 ± 21	356 ± 29	—	1634 ± 207
Notes: Data presented in g·t^−1 Al.

Table 2. Ratios of CPM and PM_2.5 to PMtotal for the Rio Tinto Alcan Sebree smelter

Location	Ratio	CPM (%)	PM2.5 (%)
FTC	(CPM or PM2.5)/PMtotal	95	97
	(CPM or PM2.5)/total PM2.5	3	4
GTC	(CPM or PM2.5)/PMtotal	55	64
	(CPM or PM2.5)/total PM2.5	11	11
Roof Vents	(CPM or PM2.5)/PMtotal	21	32
	(CPM or PM2.5)/total PM2.5	56	85
Notes: GTC/FTC and roof vent results are respectively based on OTM27/28 and M14* sampling methods. PMtotal represents the total PM emissions for each sector (FTC, GTC, and roof vents), whereas “total PM2.5” is the sum of PM2.5 emissions for all of these sectors.

The difference between GTC and FTC ratios is also worth noting (Tables 1 and 2). GTC stacks emit 3–5 times more PM than FTC stacks as each treatment unit is linked to completely different processes. The FTC recovers fumes from the anode baking furnace (ABF), where green anodes from the paste plant – where the anode paste (pitch and coke) is crushed, ground, and mixed – are baked. The fumes are then scrubbed with fresh alumina. Considering the ABF process, several other gaseous species than dust are expected, such as polyaromatic hydrocarbon, VOC, HF, and SO₂.23 It is therefore not surprising that CPM accounts for 95% of PM_total. As for potrooms, where electrolysis takes place - electric current passes through a carbon cathode and a prebaked anode (see above) to reduce alumina into molten aluminum. Process gases from cells are collected and dry-scrubbed with alumina at the GTC. Bath components, alumina, and carbon from consumed anodes are the main sources of primary PM, in a larger proportion than at the FTC. At the GTC exhaust, SO₂ (∼ 120 ppm) and HF, in a smaller proportion (< 1 ppm), are precursors of CPM.

Part 1.2: Emissions from Roof Vents

At first, the M14* and M14/OTM27/OTM28 methods were tested at the roof vents (data not shown), but no statistical differences arose. This was expected because SO₂ concentrations in the potrooms are rather diluted (<2 ppm) and hence should not cause any artifacts. Emission coefficients from the M14* method yield results approximately 10 times higher than at the GTC. By far, roof vents are undeniably the primary contributors (85%) of total PM_2.5 released by an aluminum smelter (GTC + FTC + roof vents, other sectors such as anode butt recycling or paste plants are marginal in contribution—data not shown); 56% of this total is CPM, still only from roof vents (Tables 1 and 2). All primary PM emissions are essentially the result of cell operation (anode changes, metal tapping, alumina feeding, etc.) and condensables are attributable to imperfect gas suction efficiency in cells. Hence, lowering PM emissions relies on increasing pot gas suction to recover primary PM emissions and gases responsible for CPM, both leaking from the pots.

To determine if the M14* method and the Rio Tinto Alcan CYCLEX candidate method are equivalent whatever the year of sampling (H₀), a statistical analysis was conducted. First, Gaussian distribution and variance homogeneity were confirmed by graphical examination of the residuals.24 The systematic and random errors (bias and precision) of the two methods were considered as the same constant because every weighing was conducted on the same kind of analytical balance. This error constant does not then impact further comparisons of the two methods. It is also worth mentioning the lack of standard or reference samples for repeat sampling runs on PM_2.5, making it impossible to determine a bias between the two methods. Therefore, a two-way ANOVA type I (data not matched) was performed. Next, how the PM_2.5 measurements at roof vents are affected was assessed by two independent variables for the main effects: year of measurements (2007 vs. 2008) and the method (M14* vs. CYCLEX). Lastly, if one factor affected the other was checked, what is called the “interaction effect” (year x method). Table 3 clearly shows that neither the interaction effect (year × method) nor the main effects have an effect on the model (P > 0.05). Also, the critical F-value (0.05, 1, 30) =4.171 indicates that H₀ cannot be rejected; that is, F(0.05, 1, 30) > F. Therefore, there is no evidence that any of the predictors are linearly associated with the response. It is also interesting to note that the interaction effect accounts for only 1.03% of the total variance. In other words, whatever the year of measurements, here it is demonstrated that the M14* and CYCLEX methods are interchangeable.

Table 3. Two-say ANOVA on PM_2.5 measuements at roof vents as affected by the year (2007 vs.2008 and the method (M14* vs. CYCLEX)

Sources	dF	SS	F-value	P
Year	1	0.0269	1.44	0.2403
Method	1	0.0103	0.55	0.4639
Year x method	1	0.0093	0.50	0.4854
Residual	30	0.5619

At this point, it should be noted that the CYCLEX method only seeks to give CPM and PM_2.5 emission factors at roof vents, but its design also allows measuring PM_total. Therefore, for PM_2.5 estimation (including CPM) in aluminum smelters not equipped with downcomers, the CYCLEX method seems suitable.

Part 2: Speciation of PM_2.5 in the Surrounding Environment of the Alma Smelter

Stations 3A and 1A (Figure 3) are respectively exposed to 10 and 6% of the winds passing through the smelter area, along with other emission sources that are located around these sampling sites (gas stations, pulp and paper mills, residential areas, farms). First, stations 3A and 1A show respective mean PM_2.5 concentrations of 13.3 and 8.2 μg·m⁻³ over the same 14-day period. Considering winds blowing from the smelter directly to either station and that 100% of the PM_2.5 comes from the smelter, station 3A and 1A measurements yield 1.3 and 0.49 μg·m⁻³, respectively, during the study period. Obviously, this calculation does not consider any other sources than the Alma smelter; however, farming heavily impacts station 3A, and cars and other urban activities affect station 1A.

Furthermore, from a toxicity perspective, determining the source profile for an aluminum smelter is worthwhile. Note that source apportionment calculations recognized by EPA (e.g., UNMIX, positive matrix factorization, or others25) were not used here because of a lack of long-term data, which these models require. As a first attempt, inorganic species in industrial versus ambient samples were measured (Table 4), and the results are expressed as PM_2.5 mass fraction (species mass on the filter/total PM_2.5 mass on the filter, in percent). From Table 4, it appears that species like Ga, Zr, Cd, and Be are neither detected at smelter stacks (FTC and GTC) nor in the ambient urban air. For Cr, Fe, Mg, Mn, Ca, Na, Li, Si, V, Ti, and Cu, there is an enrichment between stack emissions and the urban stations that suggests other sources (e.g., transportation, pulp and paper mills, etc.). For all other species, in particular Al, Pb, and F, mass fractions in the PM_2.5 filters collected at stacks are higher than those collected in the ambient urban air, which clearly indicates emissions from the smelter to the surrounding environment.

Table 4. PM_2.5 mass fraction of various inorganic species (species mass on the filter/total mass of PM_2.5 on the filter—in percent) at the smelter stacks (FTC and GTC) and in its surrounding environment (urban stations 1A and 3 A)

	Mass Fraction (%)
Species	FTC	GTC	Ambient Station 1A	Ambient Station 3A
Cr	<0.01	0.04	0.07	0.18
Fe	0.05	0.16	0.46	1.18
Ga	<0.01	<0.01	<0.01	<0.01
Mg	0.01	0.02	0.17	0.49
Mn	<0.01	<0.01	0.04	0.06
Ti	<0.01	<0.01	0.03	0.11
V	<0.01	0.01	0.02	0.02
Zn	0.29	0.25	0.12	0.49
Zr	<0.01	<0.01	<0.01	<0.01
Si	<0.01	<0.01	<0.01	4.34
Pb	0.24	<0.01	<0.01	<0.01
Li	<0.01	<0.01	0.01	0.01
Ni	0.03	0.08	0.03	0.05
Al	0.08	13	1.74	4.05
Be	<0.01	<0.01	<0.01	<0.01
Ca	0.03	0.08	0.74	3.61
Cd	<0.01	<0.01	<0.01	<0.01
Cu	0.01	0.22	0.29	4.01
Na	0.20	3.7	1.97	4.77
F	0.05	2.9	0.39	0.40
Total	0.98	20	6.1	24
Others	99	80	94	76

Unfortunately, elemental and organic carbon contents as well as nitrates, sulfates, and ammonium were not determined on urban filters for technical reasons. Further analyses will be necessary to better assess the effects of the smelter on ambient concentrations of these species.

Secondly, SEM-EDX tests allowed us to pinpoint some specific species released or used in aluminum smelters. Thus, from a qualitative point of view, it was determined that fine particles collected in the surrounding environment contained particles related to the electrolytic bath (chiolite, cryolite, sodium aluminum tetrafluo-ride, fluorinated alumina). At the same time, the above speciation results were confirmed in that most EDX analyses reported other types of particles (i.e., not smelter related). Indeed, Figure 4, a and b, shows EDX diagrams that match quite well for typical bath species in an aluminum smelter (same-intensity F and Na peaks and a higher Al peak), although Figure 4c shows particles that do not originate from the smelter (high Ca content and no Na).

Figure 4. SEM-EDX on GTC and urban station filters. Picture of spotted area on filters and diagram of elemental determination. EDX analysis on (a) the GTC filter, (b) the urban station 1A filter, and (c) the urban station 1A filter.

Lastly, preliminary isotope measurements demonstrate that this method could be used to track aluminum smelter PM_2.5 emissions. Stable and radiogenic isotopes ¹³C/¹²C (δ¹³C = −26.3 ± 0.2), ⁸⁷Sr/⁸⁶Sr (0.71847 ± 0.00002), and ¹⁴³Nd/¹⁴⁴Nd (0.512197 ± 0.000015) cannot be distinguished from terrestrial plants (U.S. Geological Survey database), typical soil dust,26 or weathering of till.27 However, with osmium and lead, results are very promising. Analyses were done after the process, first on anodes (main carbon raw material in the electrolytic process), then at the GTC stack (point of emission of the electrolysis), and lastly in the surrounding environment (i.e., at field station 3A) (see Figure 3). In anodes, the ratio ¹⁸⁷Os/¹⁸⁸Os =2.393 ± 0.005 is significantly different from that of usual anthropogenic sources,28 whereas ratios for ²⁰⁶Pb/²⁰⁴Pb =18.759 ± 0.005, ²⁰⁷Pb/²⁰⁴Pb = 15.625 ± 0.005, ²⁰⁸Pb/²⁰⁴Pb =38.542 ± 0.007, and ²⁰⁶Pb/²⁰⁷Pb =1.201 ± 0.002 are comparable with U.S. aerosol emissions for plant carbon sources originating from the Mississippi Valley29 (Figure 5). This means that the anodes used at the Alma facility own a specific isotopic footprint once coupled with Os and Pb (Figure 6). To evaluate smelter emissions contribution to the ambient air at station 3A, the Os ratio is considered as a mixture of smelter emissions and other sources. Considering eq 1,30

Figure 5. Lead isotope diagram for samples in the carbon raw material (anode), at the stack output (GTC), and in the surrounding environment of the smelter (station) compared with North American aerosols. Adapted with permission from Poirier.29 Copyright 2006 Elsevier, Earth and Planetary Science Letters.

Figure 6. Osmium-lead isotope diagram for samples in the carbon raw material (anode), at the stack output (GTC), and in the surrounding environment of the smelter (station).

(1)

where R _PM2.5 is the ¹⁸⁷Os/¹⁸⁸Os ratio at station 3A (0.23 measured). R _sources is the sum of the R _i·X _i ¹⁸⁷Os/¹⁸⁸Os ratios for all other sources, from usual anthropogenic sources such as catalytic converters (0.17531) to natural ones. In this case, one should note that the natural isotopic signature of Os is not considered. Indeed, the accepted value of 1.432 does not make a consensus because this value is not site specific.33 Therefore, only the most common R _sources ratio in the literature were used—catalytic converters. R _smelter is the same Os ratio for smelter emissions (2.11 measured). X _sources, which is actually equal to (1 – X _smelter), and the unknown term X _smelter are the contributing PM_2.5 factors for Os ratios. With this kind of preliminary approach, the calculation yields a contribution, from the Alma smelter, of approximately 3% at urban station 3A.

CONCLUSIONS

Part 1: Quantification of PM_2.5 Emitted by the Sebree Smelter

Part 1.1: Emissions from Stacks

All of the experience gained at Sebree Works led Rio Tinto Alcan to consider EPA OTM27 and OTM28 methods as the best approaches to quantify CPM, PM_2.5, and PM_total. Indeed, as mentioned in earlier studies,34 the wet impinger approach overestimates CPM (EPA 202), especially for aluminum smelters, because of SO₂ concentrations, whereas this is not the case with dry impingers (OTM28).

Part 1.2: Emissions from Roof Vents

For aluminum smelters without downcomers or for other industries with high ventilated building surfaces, the Rio Tinto Alcan CYCLEX candidate method is, to the best of the author's knowledge, the only way to achieve PM_2.5 (including CPM) emission coefficients. It is also essential to consider a representative location of all of the potrooms during CYCLEX sampling campaigns and to measure over a full potroom cycle length.

Part 2: Speciation of PM_2.5 in the Surrounding Environment of the Alma Smelter

This study provides a first but incomplete speciation of compounds released by an aluminum smelter to the surrounding environment. As expected, considering the electrolytic process, Al, Pb, and F are the main elements in PM_2.5 originating from a smelter. Nevertheless, the carbon fraction (organic and elemental) and the anionic species are unknown; they require further work to focus on the source apportionment approach by species. A technique such as SEM-EDX is an interesting approach to qualitatively discriminate fine particles from one source to another, but it demands a thorough knowledge of source particle signature. On the contrary, the isotopic approach, especially with the ratio ¹⁸⁷Os/¹⁸⁸Os, reveals promising results to easily track PM_2.5 smelter emissions in the ambient air because the isotopic ratio of the targeted source appears to be unique. In the future, Rio Tinto Alcan will focus its efforts on this technique.

ACKNOWLEDGMENTS

The author thanks Claude Perron (ARDC), Jacques Bélanger and Rock Emond (Alma Works), Ghassan Dugaish and Jamie Coomes (Sebree Works), and Paul Siegel and his team (Air Quality Services, Kentucky) for sampling and technical support. Rio Tinto Alcan also acknowledges Professor André Poirier from UQAM-GEOTOP for all of the work with isotopes. Rio Tinto Alcan also gives special thanks to Yvan Lavoie and Jacques Labrie (now retired from Rio Tinto Alcan) for the first works on the PM_2.5 program.