MEASUREMENTS OF PERFLUOROCARBON EMISSIONS FROM NORWEGIAN ALUMINUM SMELTERS

Transcription

1 1 MEASUREMENTS OF PERFLUOROCARBON EMISSIONS FROM NORWEGIAN ALUMINUM SMELTERS Halvor Kvande, Helge Nes and Lars Vik 1 Hydro Aluminium Metal Products Division, N-0240 Oslo, Norway 2 Elkem Aluminium Mosjøen, N-8655 Mosjøen, Norway Abstract CF 4 gas emissions from all potlines at the seven Norwegian aluminum smelters were analyzed by use of a portable photoacoustic gas monitor. A total of 8 different types of point fed prebake (PFPB) cells and 6 types of vertical stud Söderberg (VSS) cells were studied. Measurements were made in the gas exhaust duct from the potlines, and for the Söderberg cells also the fugitive gas emissions from the rooftop were measured. The results were found to be in good agreement with the corresponding series of measurements done in Norway in 1992/1993. Introduction In 1997 the Norwegian aluminum producers entered into a voluntary agreement with the Norwegian Government, through the Ministry for Environmental Protection, to reduce the CO 2 - equivalent gaseous emissions from their smelters. The goal was then to reduce these emissions, including CO 2, CF 4 and C 2 F 6, by 50 % from 1990 levels by 2000 and by 55 % from 1990 levels by Several countries have now entered into similar programs (1). Furthermore, the Kyoto protocol from 1997 includes the above-mentioned gases for potential emissions reductions and has established the reduction targets. Due to the nature of the aluminum production process, where CO 2 is formed from the consumable carbon anodes, little gain is possible there. Thus, the major reduction in greenhouse gas emissions has to be made in the s. As a part of the obligations included in the voluntary agreement, the Norwegian aluminum industry has now made new measurements on all the different types of cells in Norway, in order to verify, and if necessary, correct the measurements made in 1992/1993 (2). Since that time, several of the Söderberg potlines have been equipped with point feeders, and it was also considered to be of interest to determine if that has had any influence on these emissions. The purpose of the present paper is to report the results of the new measurements and to make a comparison with the previous Norwegian data. Furthermore, the results will be compared with other data published in the open literature. A recent IAI report (3) states that further work on emission measurement and uncertainty analysis should be pursued, especially on vertical stud Söderberg cells. This has been a significant part of the present measurements. A recommendation of the best data will now be made to the Norwegian Pollution Control Authority. Thus, these data make it possible to obtain an accurate inventory of the quantities of CF 4 generated annually by the Norwegian aluminum industry. The formation of CF 4 and C 2 F 6 during anode effects and their importance as greenhouse gases, and thus for global warming, have been discussed in several papers (1-6). This will therefore not be repeated here and the readers are referred to these for additional background information. Measurement Technique The measurements of s at the smelters were performed to account for both emissions captured by the reduction cell hooding and extracted by the fume exhaust duct, and also the fugitive emissions released into the potroom atmosphere. This was done for all Söderberg potlines, while the measurements for prebake cells showed that the fugitive emissions were negligible. This agrees with our previous finding (7) that the hooding capture efficiency for these cells is high, between 98 and 99 %. Experimental A photo-acoustic gas monitor, Bruel & Kjær, model 1302, was used in these measurements, with an optical filter for micrometer IR frequency. For further details on this equipment the

2 2 paper by Berge et al. (2) is referred to. The instrument works in the infrared range, which means that it can record gases that absorb infrared light. This in turn means that if H 2 O, SO 2 and CH 4 are present, they may interfere with the CF 4 measurements. These gases must therefore be removed before the gas sample enters into the instrument. An ascarite filter removed SO 2 and CH 4 from the gas, while the water vapor was removed by the use of a tube with blue gel and a copper spiral embedded in solid carbon dioxide. The company Innova in Copenhagen, Denmark, which is a daughter company of Bruel & Kjær, did the calibration of the instrument. The calibration was made with respect to any interference with the above-mentioned gases, so that their presence was corrected for in the calculation of the results for CF 4. The accuracy of the instrument is! 10 %. This is the instrumental accuracy under static conditions, that is, the accuracy of continued measurements on a source of CF 4 of known and unchanging concentration. The lower detection limit was 0.1 ppm when pure gas was measured. In the previous measurements (2) the data obtained by use of this instrument was compared with similar measurements by use of a mass spectrometer and a Fourier Infrared spectrometer. Good conformity between these three methods was obtained at gas concentrations up to about 40 ppm. Sampling The gas sample was pumped out of the gas duct, filtered to remove dust and then passed through an air trap. Then the sample was led through an ascarite absorption tube to remove SO 2 and a copper tube cooled to about 70 o C in solid carbon dioxide to remove any remaining humidity. The monitor then analyzed one gas sample every 1.5 minutes. The actual sampling phase of these 1.5 minutes was about 10 seconds, while the rest of the time was spent doing the IR analysis and other pumping and flushing work. In addition, also H 2 O, SO 2 and C 2 F 6 were analyzed and recorded. The duration of the measurements was between two and twelve hours. In modern prebake potlines the anode effect frequency is low, typically between 0.02 and 0.10 anode effects per cell-day, so that anode effects had to be provoked by stopping the alumina feeding for some selected cells. For Söderberg cells this was also done, but not when the fugitive gas emissions were measured. Anode effect data from the cell computers were recorded, with particular emphasis on their duration or length. The duration of each anode effect was taken as the total time when the cell voltage was above 8 V. Duct volumetric gas flow measurements were made at the same time as the measurements for all cell lines in this study. The anode gas volume was then measured and the normalized volume (corrected for temperature and pressure and given in Nm 3 /h) was calculated for all measurement positions. The photo-acoustic gas monitor measured only concentrations and the gas volume was therefore needed to calculate the total gas emissions. The instrument did not record any perfluorocarbon gases during normal cell electrolysis. Results The largest weakness with the use of the photo-acoustic analyzer is of course that it measures only every 1.5 minutes. Due to the nature of the measurements, the period when the instrument recorded a higher than normal concentration of CF 4, was considerably longer than this. Typically, the anode effect duration in the present study was about 4 minutes, while the recording of CF 4 concentrations could last for twice that time, and even longer in some cases. Thus, the number of measurements for each anode effect could vary from three to more than ten. It is then a question of how accurate the instrument can calculate the average of a signal that can change from zero to a maximum in a fraction of a minute. Intuitively, we do not think it would result in a bias, but it could introduce a measurement error. The uncertainty in the results due to this factor is unknown at this point. The CF 4 concentration was calculated by taking the average value of all the readings during the period when the instrument showed that this gas was present. This average value was divided by the recorded duration of the CF 4 readings and was then multiplied by the gas volume to calculate the amount of CF 4 emitted for each anode effect. The emission per anode effect-minute was calculated by dividing the total emissions by the actual average anode effect duration, as recorded by the cell computer. In a few cases two anode effects occurred simultaneously, and the curves for CF 4 concentration as a function of time then overlapped, sometimes with two distinct peaks. The total gas concentration of CF 4 was then divided by the number of anode effects to calculate the average CF 4 gas emission per anode effect. We also made measurements of the fugitive gas emissions from Söderberg potlines. The instrument was then positioned so that it recorded 12-hour rooftop concentrations. There was considerable moment to moment variability. The results were expected to be somewhat uncertain, because the measured rooftop concentrations were dependent on several factors, mainly the distance from the cell having an anode effect, but also the sampling location and the variable vent flow in the potroom. Particularly the distance from the sampling position to the cell on anode effect is crucial here. The optimum would be if the cell just right underneath the sampling position had an anode effect, but even then one cannot be sure to collect all the CF 4 emitted. The emissions from all other cells would be more diluted on their way up to our sampling position.

3 3 In these cases we believe that we can never measure a too high value for the s, so we have chosen here to report the highest value we found for each potline. This concentration was then applied to the measured vent flow through the rooftop to calculate the amount of CF 4 per unit time. Added to the regular exhaust duct measurements, these data gave the total CF 4 emissions from this type of Söderberg cells. Similar rooftop measurements have been attempted also for the prebake potlines, but showed no detectable concentrations. Even if this does not justify the conclusion that there are no fugitive emissions from prebake cells, it is clear that these must be considerably lower than from Söderberg cells. The results for the Söderberg cells are shown in Table 1. Table 1. s from the exhaust duct and fugitive emissions from the rooftop for vertical stud Söderberg (VSS) cells. Potline no. Amperage (ka) Exhaust duct emissions (kg CF 4 /AE minutes) Rooftop emissions (kg CF 4 /AE minutes) Total emissions, exhaust duct plus rooftop (kg CF 4 /AE minutes) Percentage from rooftop emissions (%) (estimated) The contribution from the fugitive emissions is seen to be fairly constant at 17! 3 % for these potlines. This value agrees well with the figure given by IPCC (3), who estimated that the emission collection efficiency for VSS cells was 15 %. In the previous Norwegian measurements (2) it was reported that approximately 11 % of the CF 4 gas escaped from Söderberg cells into the potroom during anode effects. At that time the approach was to determine the average value of the anode effects of the cells in the actual positions underneath the ventilator at the measuring point. The measured s are given in Tables 2 and 3 for the prebake and Söderberg potlines, respectively. The potline amperage is given, together with data for the number of anode effect minutes per cell-day and the per metric ton (mt) of Al produced, which is also given as metric ton of Al produced per AE minutes per cell-day. For the prebake potlines the slope values shown in the right hand column in Table 2 vary between 0.10 and 0.19 kg CF 4 /mt Al/AE minutes/cell-day, with an average value of 0.16! 0.03 kg CF 4 /mt Al/AE minutes/cell-day. For the Söderberg potlines the values in Table 3 are generally lower, but with larger variation, between 0.05 and 0.19, with an average of 0.11! 0.05 kg CF 4 /mt Al/AE minutes/cell-day. Table 2. Data for s measured in 1999/2000 from eight different Norwegian prebake (PFPB) potlines. The potlines with amperages above 180 ka have side-by-side cells. Potline no. Amperage (ka) Anode effect duration per day (AE minutes/cell-day) (kg CF 4 /mt Al) ! 0.05* ! 0.03* ! 0.05* ! 0.05* ! 0.05* ! 0.05* ! 0.06* ! 0.04* Average: 0.16! 0.03 * We assume a relative standard deviation of! 30 % on the basis of the previous Norwegian measurements (2).

4 4 Table 3. Data for s measured in 1999/2000 from six different Norwegian Söderberg (VSS) potlines. Fugitive emissions are included also in the data expressed as kg CF 4 /mt Al. Potlines no. 2, 4 and 5 do not have point feeders. Potline no. Amperage (ka) Total anode effect duration per day (AE minutes/cell-day) (kg CF 4 /mt Al) ! 0.04* ! 0.03* ! 0.01* ! 0.01* ! 0.03* ! 0.02* Average: 0.11! 0.05 * We assume a relative standard deviation of! 20 % on the basis of the previous Norwegian measurements (2). This value is lower than for prebake cells because of the larger number of anode effects measured for Sõderberg cells. Comparison with IPCC Slope Coefficients IPCC guidelines (3) for comparison of s from primary aluminum producers recommend multiplying the number of anode effect minutes per cell-day by a slope factor to obtain kg CF 4 per unit (mt) of aluminum production. A comparison with literature data is given in Table 4 for prebake cells. It is seen that these potlines gave values that agree reasonably well. The recommended IPCC value of 0.14! kg CF 4 /mt Al/AE minutes/cell-day is based on the best available data in the literature. The average value from the present work, 0.16! 0.03, is in good agreement with the IPCC value, but it should be remembered that the relative uncertainty in each of the present data is rather large (! 30 %). The present measurements thus show that there is considerable variability in s from smelter to smelter. The main question is then if the data from various smelters still are the same, within the uncertainties given. The IPCC Tier 2 approach is to have an accepted value for use by all smelters of similar technology. Intuitively, it may make sense to suggest that a common value should be determined from the literature, which might be used by all smelters having prebake cells with point feeders. In the present measurements we have chosen this approach. Table 5 shows the corresponding results from the recent measurements on vertical stud Söderberg cells. Also here the results are in reasonably good agreement. The relative uncertainty in the present data is estimated to be! 20 %, on the basis of the measurements by Berge et al. (2). However, as seen from Table 2, values in the range from 0.05 to 0.19 kg CF 4 /mt Al/AE minutes/cell-day have been determined in the present work. Three of the potlines do not have cells equipped with point feeders, but this does not seem to cause any systematic deviation in the data. The present average value is well within the recommended IPCC value for vertical stud Söderberg cells. In Tables 4 and 5 we have also calculated the average values from the Norwegian measurements in 1992/1993, in 1996/1997 and from the present ones. The same instrument and the same measurement technique were used in all cases. The overall agreement between the results is good, and this therefore seems to be the best way of treating the results. It is seen that the deviation of the average values from those obtained in the 1999/2000 measurements is small and is well within the standard deviation. Table 4. Comparison of literature data for s (IPCC Slope Coefficients) from point-fed prebake (PFPB) cells. Reference s Comments Berge et al. (2) Six Hydro Aluminium potlines measured in 1992/93 Leber et al. (4) Average of seven literature values Marks et al. (5) 0.13 Nine Alcoa potline measurements Marks et al. (6) 0.16 and 0.22 Two US potline measurements IPCC (3) 0.14! Recommended IPCC value in 1999 for CWPB* cells Present work 0.16! 0.03 Eight Norwegian potlines in 1999/2000 Present work 0.15! 0.02 Average of the Norwegian results from 1992/1993 and the present data from 1999/2000 *CWPB here means Center Worked Prebake Cells, but it does not necessarily imply that they have point feeders.

5 5 Table 5. Comparison of literature data for s (IPCC Slope Coefficients) from vertical stud Söderberg (VSS) cells. Reference s Comments Berge et al. (2) Four Hydro Aluminium potlines measured in 1992/93 Leber et al. (4) Average of five literature values Marks et al. (5) One Alcoa potline measurement Marks et al. (6) 0.12 One US potline measurement IPCC (3) 0.068! 0.02 Recommended IPCC value in 1999 Present work 0.11! 0.05 Six Norwegian potlines in 1999/2000 Present work 0.10! 0.02 Average of the Norwegian results from 1992/1993, 1996/1997 and the present data from 1999/2000 Discussion The Anode Effect Over-Voltage method An alternative to the slope method for calculation of the relevant emission factor is to use the so-called Anode Effect Over-Voltage method (3,8). The Anode Effect Over-Voltage integrates the fluctuation in voltage during an anode effect. Thus, it is the extra cell voltage, in this case above 8 V, caused by anode effects, when averaged over a 24-hour period. It should certainly not be confused with the usual electrochemical overvoltage term. Here, it has the unit of mv day per cell-day. In the IAI report (3) it was erroneously reported as mv/day. In one of the present measurements (potline no. 6 in Table 2) the duration of the anode effects was measured manually in order to make these results directly comparable to the other data from the slope method. This was necessary because the process control system there included a certain waiting period after the anode effect is quenched, and thus the reported anode effect duration was too long. Anode effects vary with respect to s, even in the same cell Significant differences in s were found in both prebake and Söderberg cells. In this respect it may seem that each anode effect is different from the previous one. The sequential lowering and raising of the anodes during automatic anode effect quenching cause short-circuiting between the carbon anodes and the aluminum metal pad. This short-circuiting is not necessarily reproducible with respect to time, which in turn results in fluctuations in the s. This can explain why the results can differ between individual anode effects in the same potline and even in the same cell, and not to mention the measurements in different aluminum smelters. This may also explain why the uncertainty in the measurements may be so large as 20 to 30 %. We are actually measuring something that is not constant from one anode effect to another. In this respect it is not surprising that the IPCC recommended values have relative standard deviations of about! 30 % for VSS cells. Effect of point feeding of Söderberg cells Some of the Söderberg potlines have been equipped with point feeders since This is the case for potlines no. 1, 3 and 6 in Table 3. For two of these potlines the results from the present measurements were identical to the previous ones, while in one case the new data were somewhat higher, but still within the given uncertainty. It is therefore concluded that the measurements do not indicate that introduction of point feeders in itself has changed the rate for Söderberg cells. It seems that the duration of the anode effect is a much more important parameter here. Difference between prebake and Söderberg cells Do Söderberg cells really emit less CF 4 than prebake cells? First, it depends upon the unit of emission. It is obviously true if we talk about emissions per AE minute, and we thus ignore the effect of the cell size. On the other hand, the data in Tables 2 and 3 clearly show the opposite if we calculate emissions only per metric ton of aluminum produced. Then Söderberg cells emit much more CF 4 (by a factor of ten from the data in tables 4 and 5). The IPCC recommends that emissions be given in kg CF 4 /mt Al/AE minutes/cell-day. Then the IPCC values indicate that prebake PFPB cells emit about twice as much CF 4 as Söderberg cells. This may be connected with frequent short-circuiting of the much less smooth working surface of the Söderberg anode during the anode effect quenching. The CF 4 concentration in the anode gases will fluctuate accordingly. On the other hand, the present data for s from prebake and Söderberg cells are equal, within their uncertainties. Thus, it is quite possible that the difference is significantly less than indicated by the IPCC values. This deserves to be studied further.

6 6 Emissions of C 2F 6 With the present experimental technique the measured C 2 F 6 concentrations from prebake cells showed distinct increases during anode effects and they were therefore considered sufficiently reliable to attempt a further evaluation. On the other hand, the C 2 F 6 data from the Söderberg cells did not change significantly and were not used here. The prebake potlines gave an average value for the C 2 F 6 emission of 0.011! kg C 2 F 6 /mt Al/AE minutes/cell-day. These measurements then indicate that 6 % of the total perfluorocarbon gas emissions during anode effects is caused by C 2 F 6. Acknowledgement The authors gratefully acknowledge the very valuable comments given to this paper by Jerry Marks of IAI. Conclusions - s from eight prebake cell types (PFPB) and six Söderberg cell types (VSS) in Norway were measured by use of a portable photo-acoustic gas monitor (Bruel & Kjær, model 1302). - The present data showed good agreement with the measurements that were made in Norway in 1992/ For the PFPB cells data in the range from 0.10 to 0.19 kg CF 4 /mt Al/AE minutes/cell-day have been determined in the present work. Within the uncertainty in the results, the data for each PFPB potline agree well, and there is then no reason to report different results for each of these potlines. References 1. Dolin, E. J., PFC Emissions Reductions: The Domestic and International Perspective, Light Metal Age, 58 (1,2) (1999), Berge, I., Huglen, R., Bugge, M., Lindstrøm, J. and Røe, T. I., Measurement and Characterization of Fluorocarbon Emissions from Alumina Reduction Cells, Light Metals 1994, Anode Effect Survey and Perfluorocarbon Compounds Emissions Survey , published by International Aluminium Institute (IAI), London, UK, 2000, 28 pp. 4. Leber, B. P. Jr., Tabereaux, A. T., Marks, J., Lamb, B., Howard, T., Kantamaneni, R., Gibbs, M., Bakshi, V. and Dolin, E. J., Perfluorocarbon (PFC) Generation at Primary Aluminum Smelters, Light Metals 1998, Marks, J., PFC Measurements from Alcoa Aluminum Smelters, Light Metals 1998, Marks, J., Perfluorocarbon (PFC) Generation during Primary Aluminum Production, Light Metals 2000, Karlsen. M., Kielland, V., Kvande, H. and Vestre, S. B., Factors Influencing Cell hooding and Gas Collection efficiencies, Light Metals 1998, Bouzat, G., Carraz, J.C. and Meyer, M., Measurements of CF 4 and C 2 F 6 emissions from Prebaked Pots, Light Metals 1996, For the VSS cells data in the range from 0.05 to 0.19 kg CF 4 /mt Al/AE minutes/cell-day have been determined in the present work. Here, this variation is larger than expected from the measurement uncertainty, and it may partly be caused by a real variation from the different sites. - For the Norwegian PFPB cells the Slope Coefficient is 0.15! 0.02 kg CF 4 /mt Al/AE minutes/cell-day, as the average value from the present and previous Norwegian measurements (2). - For the Norwegian VSS cells the corresponding Slope Coefficient is 0.10! 0.02 kg CF 4 /mt Al/AE minutes/cell-day, as the average value from the present and previous Norwegian measurements (2). - The IPCC Slope Coefficients calculated from the present work agree very well with data from the literature and with the recommended IPCC values.