Research Concentration Variability of Halocarbons over an Electronics Industrial Park and Its Implication in Compliance with the Montreal Protocol CHIH-CHUNG CHANG,† GIUNN-GUANG LO,† CHENG-HSIUNG TSAI,‡ AND J I A - L I N W A N G * ,† Department of Nuclear Science, National Tsing-Hua University, Shin-Chu 300, Taiwan, and Department of Chemistry, National Central University, Chungli 320, Taiwan
This work investigated fugitive emissions of anthropogenic halocarbons in a semiconductor and electronics industrial park in Taiwan using both flask and in-situ measurement methods. Large concentration variabilities in methylchloroform, trichloroethene, and tetrachloroethene suggested substantial usage and emissions in the industrial park. While the variability of CFC-113, CCl4, and CFC-11 was rather small using the flask sampling technique, the in-situ method with its higher frequency, however, showed significantly larger variability arising from observing periodic emission episodes, which were highly correlated with wind direction and topography of the park.
Introduction The semiconductor and electronics industries have grown significantly in the past decade worldwide due to the strong demand for information products. Production of integrated circuit chips, electronic circuit boards, and various other related products requires a series of cleaning processes in which a number of chlorinated organic solvents are used and ultimately vented into the atmosphere with or without abatement. These chlorinated solvents once involved ozonedepleting substances, e.g., trichlorotrifluoroethane (CFC-113) and methylchloroform, which were commonly used as deflux agents for cleaning circuit boards. Because of the implementation of the Montreal Protocol (1) and subsequent amendments (2) to reduce and ultimately cease the production of these compounds by 1996, applications of CFC-113 and methylchloroform were largely substituted by other compounds. Common solvents found in the cleaning process include methylene chloride, dichloroethane, trichloroethene, tetrachloroethene, and other non-chlorine-containing solvents such as 2-propanol and acetone. Although some of these compounds still contain chlorine, their double bonds or hydrogen atoms are easily subject to hydroxyl radical addition or abstraction in the atmosphere. This removal process renders their atmospheric residence time relatively * Corresponding author e-mail:
[email protected]; fax: +886 3 4277976. † National Tsing-Hua University. ‡ National Central University. 10.1021/es001894q CCC: $20.00 Published on Web 07/12/2001
2001 American Chemical Society
short; thus, the fraction that enters into the stratosphere is small (3, 4). Nevertheless, reckless use or emissions of these chlorinated solvents could still cause concerns on human health from long-term exposure in either the working environment or the residential areas near the factories. Investigation of emission was conducted in an industrial park densely packed with semiconductor or electronics manufacturers with approximately 300 of these types of factories concentrated in an area of only 4 km2. The revenue generated from this park amounts to 8% of Taiwan’s GNP annually as compared to only 0.5% a decade ago. Although this industry is a success from an economic point of view, the burden it creates on the environment resulting from fugitive emissions also has become a serious treat to the environment and the public heath. In addition, since the EPA of Taiwan pledged to adhere to the Montreal Protocol and implement the mandates, intensive measurement conducted in an industrial park of substantial size can shed light on the effectiveness of the implementation of the treaty in Taiwan. In this study, ambient air was measured for selected chlorinated compounds to assess their usage and emissions from spatial and temporal concentration variability by flask sampling and in-situ monitoring, respectively.
Experimental Section Hourly measurement was made by an in-situ GC with each sampling aliquot of 150 mL collected over a duration of 15 min. The sample was drawn through a multisorbent trap kept at 30 °C for preconcentration and subsequently heated within seconds to 250 °C for flash desorption on to a DB-1 column (J&W, Folsom, CA; 60 m × 0.32 mm; df ) 1.0 µm). This column was housed in a Varian 3400 GC equipped with an electron caption detector. Design of the concentrator and the control hardware/software appear in more detail in our previous publication (5). Precision was better than 2% (1σ or 1 standard deviation, SD) estimated from continuous injection of a clean ambient air sample (n > 27). This in-situ GC/ECD was placed in a trailer with an air inlet extruding the trailer top. Figure 1 shows the map of the industrial park and the location of the trailer. A wind sensor was in operation along with the in-situ GC during the 11 days of continuous measurement from May 7 to May 18, 2000, to obtain wind speed and direction every 5 min. The status of the in-situ GC and the resulting chromatograms can be viewed and remotely controlled from the central laboratory via modems and phone lines. Figure 2 displays a typical chromatogram from the in-situ measurement. In addition to the in-situ measurement, 42 flask samples were also randomly collected in the park with 3-L stainless steel canisters (SilcoCan, Restek) within a 30-min period on July 25, 2000. The samples were brought back to the lab and analyzed by a Varian 3800 GC/MS for chlorinated compounds. The same type of column as for the in-situ GC, i.e., DB-1, was used in this system. Blank runs were made to confirm the cleanness of the canisters before shipping out for sampling. Earlier results from two sampling collections, i.e., 25 samples conducted in the same park in March 1997 and 65 coastal samples collected around the island in February 1997 are included in this study for investigating the evolution in concentration distribution over time. Samples from these two collections were analyzed by GC/ECD (Varian 3400 CX) with the same built-in cryogenic preconcentration design as the GC/MS system. Precision from both cryogenic GC systems VOL. 35, NO. 16, 2001 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
3273
FIGURE 1. Sampling locations within and around the industrial park. Flask sampling sites are indicated by solid circles, and the in-situ site is indicated by the gray box, on which a cross is drawn to divide the studied area into quarters to be discussed in connection with wind direction in the text. was better than 3% based on repeatedly injecting a clean air sample (n ) 10). Absolution calibration was made by preparing a series of standards diluted from pure compounds (6).
Results and Discussion Figure 3 shows the statistics of the seven species from two sample collections in the park with 25 samples collected in March 1997and 42 samples collected in July 2000, respectively. The concentration variability for CFC-113 and CCl4 in either collection was the smallest among the seven measured species. When excluding the outlying values of upper and lower 10 percentiles, their 1σ relative standard deviation (RSD) became 2.8% and 2.8% for the 2000 collection and 2.8% and 3.9% for the 1997 collection for CFC-113 and CCl4, respectively. Both CFC-113 and CCl4 were primarily used as deflux or degreasing agent in the electronics industry in the preprotocol era. In the post-Montreal Protocol era, however, electronic products are not supposed to be cleaned with CFC-113 to avoid heavy tariffs under the Montreal Protocol. Instead, other alternative agents, such as water- or hydrocarbon-based solvents, have replaced CFC-113 and CCl4 in the cleaning process. Because of the long global average lifetimes of about 90 and 40 yr for CFC-113 and CCl4, their atmospheric distribution is found rather uniform in many parts of the world years after their production was halted. On a global scale, this phenomenon was revealed in the slowdown in the atmospheric growth rate accompanied by 3274
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 35, NO. 16, 2001
the gradual reduction in N/S gradient (7, 8). On a local or regional scale, it is reflected by the very tight concentration ranges across the surveyed areas including urban or suburban environments (9, 10). For instance, our previous flask sample collection in metropolitan Taipei showed a 1σ RSD of 1.41% for CCl4 and 4.8% for CFC-113 for all 109 samples, only slightly higher than their instrumental precision, suggesting that usage and therefore emissions in a nonindustrial environment were quite limited (11). After the outlying values were taken into account, the deviation of CFC-113 and CCl4 for the two industrial park collections became significant larger than that of the urban collection (9.7% for CFC-113 and 11.7% for CCl4 for the 2000 collection), suggesting the existence of minor release. Although random sampling with adequate spatial coverage can provide an overall prospect about the relative usage or emissions for these regulated compounds as shown in Figure 3, this somewhat similar to a “snapshot” result may not be entirely representative of the all-time production routine. However, it becomes more informative and conclusive when combined with time series observation. In-situ on-site monitoring of anthropogenic halocarbons at background or quasi-background locations to examine the evolution of these compounds from the aspect of decline in growth rates or variability has been performed with long records by several research groups (7-9, 12-15). Nevertheless, in-situ measurement performed within a potential source area or for a particular industry to inspect fugitive
FIGURE 2. Typical chromatogram from the in-situ hourly measurement.
TABLE 1. Concentration Variability Represented by 1σ RSD for Data Points Lie in the Range of the 10th and 90th Concentration Percentile coastal samples in 1997a (%)
park samples in 1997b (%)
park samples in 2000c (%)
in-situ data in 2000d (%)
1.5 3.8 3.4 1.2 2.2 75.4 69.1
2.8 3.9 2.1 0.8 34.3 85.2 108.3
2.8 2.8 9.6 2.6 15.6 54.0 21.0
6.9 7.8 8.6 14.6 9.0 46.1 23.5
CFC-113 CCl4 CFC-12 CFC-11 CH3CCl3 C2HCl3 C2Cl4 a
65 samples.
b
25 samples. c 42 samples.
d
248 aliquots.
CFC emissions and hence to evaluate the degree of compliance with the protocol obligations has seldom been reported. As a result, continuous hourly measurement was conducted for 11 days covering weekdays and a weekend, as shown in Figure 4 for seven halocarbons for capturing periodic concentration spikes. Instead of showing absolute concentrations in Figure 4, the relative concentrations normalized to the minimum concentrations encountered over the 11day period are illustrated for easy comparison between compounds for their excessive abundance due to local release. To broaden the comparison base, concentration variability for this in-situ measurement was compared with that for the flask sample collections in the park and background environment (see Table 1). Note that the concentration variability for CFC-113 and CCl4 from the insitu approach was significantly larger than that observed from the flask sampling method. Furthermore, upon closer examination of the concentration spikes in the in-situ observation, it was found that the spikes of CFC-113, CCl4, and other halocarbons were in close relationship with wind speed and direction, which will be discussed later in the text.
FIGURE 3. Concentration distributions of halocarbons from two separate sample collections: (a) 42 samples collected in July 2000; (b) 25 samples collected in March 1997. Whiskers define the outlying 5% of the data. The box represents 10-90% of the data. The median is the line within the box. VOL. 35, NO. 16, 2001 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
3275
FIGURE 4. Relative concentration variation for (a) CCl4, (b) CFC-113, (c) CFC-12, (d) CFC-11, (e) CH3CCl3, (f) CHCldCCl2, and (e) CCl2dCCl2. Relative concentration is calculated as the absolute concentration for each measurement aliquot divided by the minimum concentration encountered during the 11 days. As a result, the on-site continuous measurement deployed in the potential source area was able to sensitively capture sudden fugitive emissions of CFC-113 and CCl4 which otherwise would likely be missed by the flask sampling method and exhibit much smaller variability. CFC-12 and CFC-11, which once accounted for about 80% of the CFC emissions worldwide (16) in the pre-protocol era, were mainly used as refrigerants in vehicles or large buildings. In this park study, both flask and in-situ data of CFC-12 exhibited significantly larger concentration variability than the 1997 park collection, as shown in Table 1. The major source of CFC-12 is air conditioners in older vehicles. Cars 3276
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 35, NO. 16, 2001
made after mid-1994 were required by the Taiwan EPA to replace CFC-12 with HFC-134A; hence, it was reasonable to believe that older cars using CFC-12 were more scarce in 2000 than in 1997. Thus, the lager variability of CFC-12 from the recent flask or in-situ measurement was inconsistent with such a replacement progress; therefore, the seasonality with the use of air conditioners was thought to play a dominant role in controlling emissions. Also used as a refrigerant but mainly used in large central conditioning system for buildings is CFC-11, which exhibited significantly larger variability with the in-situ measurement than the flask measurement as result of observing large
FIGURE 5. Concentration vs wind direction for (a) CCl4, (b) CFC-113, (c) CFC-12, (d) CFC-11, (e) CH3CCl3, (f) C2HCl3, and (g) C2Cl4. Panel h shows wind speed vs wind direction. concentration spikes (see Figure 4d). Similar to CFC-12, the variability of CFC-11 from either in-situ or flask sampling was also larger than the 1997 survey. We suspect that this larger variability in recent measurement for either CFC-12 or CFC-11 could be attributed to the timing of sampling in that most of the air conditioners were in use during the measurement period in mid-May for in-situ measurement
and in July for flask measurement. By contrast, the park measurement conducted in winter 1997 resembled a “baseline” condition when most of the compressors in the park were not in operation. Although we cannot rule out another possibility that more usage of CFC-12 and -11 has occurred since 1997 due to the rapid growth of factories in the park, this possibility is less likely because newer models of VOL. 35, NO. 16, 2001 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
3277
refrigeration units are not compatible with these two compounds. Figure 4e shows the variability of CH3CCl3 from the insitu measurement. This compound was commonly used as a cleaning solvent in the manufacturing process. Under the Montreal Protocol, CH3CCl3 production was no longer allowed. Because of its relatively short lifetime of about 5 yr, mainly due to reaction with hydroxyl radical, its average global background concentrations declined more rapidly than longer-lived CFCs from its peak Northern Hemisphere concentrations of about 140 pptv in 1991 to the current level of about 70 pptv in 2000 (extrapolated from refs 12 and 13). The cease of CH3CCl3 production is consistent with our recent flask and in-situ measurements, which showed that the variability of CH3CCl3 has reduced substantially when compared with the 1997 park sampling (see Figure 3). Dramatic reduction in variability of CH3CCl3 was also observed in two urban collections conducted 2 yr apart (11, 17). Although the five halocarbons discussed above (i.e., CCl4, CFC-113, CFC-12, CFC-11, and CH3CCl3) are subject to the complete phase-out by 1996 under the 1992 Copenhagen amendment (2), our measurement revealed that their minor emissions or usage still persist. This could be attributed to the ongoing consumption of the remaining stockpiles. Customs records show that the imports have dropped to zero since 1992. However, the possibility of no shortage of CFC supply by smuggling from nations that are exempted from the Protocol cannot be ruled out. Figure 4, panels f and g show in-situ data of CHCldCCl2 and CCl2dCCl2. These two species have low ozone-depletion potentials and therefore are not regulated by the Montreal Protocol. However, voluntary reduction in emissions of these two compounds occurred in the United States and Europe in light of toxicity concerns. Because of their relative high reactivity arising mainly from reacting with HO radicals and, to a lesser extent, Cl radicals (τ ≈ 7 days for CHCldCCl2; τ ≈ 4 months for CCl2dCCl2), their background mixing ratios are low (4, 18-21). Hence, spatial concentration variability is more pronounced than longer-lived CFCs due to large concentration gradient between source and background environments, which explains the extremely large variability of these two compounds in the in-situ measurement shown in Figure 4f,g. Decreasing trend in consumption in Taiwan was also noted for both compounds from the peak value of about 8 kt in 1996 decreasing to 3 kt in 1999 for trichloroethene and about 1.0-0.4 kt for tetrachloroethene. This decline in consumption is in agreement with our measurement results. In comparison with the 1997 park collection, the variability of these two compounds from either the insitu or the 42 flask measurements became substantially reduced, accompanying by lower concentration levels (see Figure 3). The precision of the in-situ system was better than 2% for all the measured compounds. Consequently, if assuming constant emission strength, any variability larger than the system’s reproducibility ought to be caused by the various degrees of dilution from sources that can be rationalized when incorporating wind direction and speed information (see Figure 5). Referring to Figure 1, the in-situ site experienced very little wind from 260° to 350° (from west to north); hence, the hourly measurements with 15-min sampling time for each aliquot captured winds mostly outside this angle range. In general, with the exception of CFC-11, for all the other halocarbons (i.e., CFC-12, CFC-113, CH3CCl3, CCl4, C2HCCl3, and C2Cl4), the in-situ site observed elevated concentrations whenever air parcels coming from north to east (about 0-90° in Figure 5) were experienced. This suggests that the existence of sources and the transport to the observation site did not allow enough mixing to dilute 3278
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 35, NO. 16, 2001
the concentrations into baseline levels. By contrast, air masses from southeast to west were generally cleaner, which mostly correspond to the low points in Figure 4, suggesting that the sources in this direction were either minor or subject to greater dilution. This is consistent with the topography of the park in that the major distribution of the plants are northeast of the in-situ site and no or few plants were scattered to the south as shown in Figure 1. The northeast winds, which picked up emissions from the most developed section of the park along its passage, consistently appeared with a very small angle span and had lower speed than winds from other directions (see Figure 5h). The slower winds allowed less dilution of emissions within the park and facilitated the buildup of halocarbon concentrations. Thus, this in-situ observation combined with wind information is useful in assessing emission characteristics associated with a certain source area, particularly when they are highly variable with wind direction or manufacturing routine. While the random flask sampling method provides an overall prospect about the relative usage or emissions for the ozone-depleting substances over a potential source area, the abrupt emission events can be more precisely recorded and better rationalized by the in-situ high-frequency measurement. The minor emissions of CFC-113 and CCl4 detected by the in-situ method combining with wind data suggest that their applications in the electronics industry were largely replaced and are therefore in reasonably good compliance with the Montreal Protocol. By contrast, the emissions for other anthropogenic halocarbons were still found to be significant. As a result, from the standpoint of complying with the Protocol for CFC-12, CFC-11, and CH3CCl3 or of health issues for CHCldCCl2 and CCl2dCCl2, measuring the concentration variability in a potential source area proves to be a useful method for sensitively determining the existence of fugitive halocarbon emissions.
Acknowledgments The authors thank the personnel who participated in air sampling and analysis. This research was supported by the National Science Council of Taiwan under Contract NSC892113-008-001.
Literature Cited (1) United Nations Environmental Programme (UNEP). Montreal Protocol on substances that deplete the ozone layer. September 16, Montreal, 1987. (2) United Nations Environmental Programme (UNEP). Report on the fourth meeting of the parties on substances that deplete the ozone layer. Copenhagen, 1992. (3) Atkinson, R. Atmos. Environ. 1990, 24A, 1. (4) Wang, C. J.-L.; Blake, D. R.; Rowland, F. S. Geophys. Res. Lett. 1995, 22, 1097. (5) Wang, J. L.; Chang, C. J.; Chang, W. D.; Chew, C.; Chen, S. W. J. Chromatogr. A 1999, 844, 259. (6) Wang, J. L.; Lin, W. C.; Chen; T. Y. Atmos. Environ. 2000, 34, 4393. (7) Fraser, P. J.; Cunnold, D. M.; Weiss, R. F.; Prinn, R. G.; Simmonds, P. G.; Miller, B.; Langenfelds, R. J. Geophys. Res. 1996, 101, 12,585. (8) Simmonds, P. G.; Cunnold, D. M.; Weiss, R. F.; Prinn, R. G.; Fraser, P. J.; McCulloch, A.; Alyea, F. N.; O’Doherty, S. J. Geophys. Res. 1998, 103, 16,017. (9) Hurst, D. F.; Bakwin, P. S.; Elkins, J. W. J. Geophys. Res. 1998, 103, 25, 299. (10) Grosjean, E.; Rasmussen, R. A.; Grosjean, D. Environ. Sci. Technol. 1999, 33, 1970. (11) Wang, J. L.; Chew, C.; Chen, S. W.; Kuo, S. R. Environ. Sci. Technol. 2000, 34, 2243. (12) Prinn, R. G.; Weiss, R. F.; Miller, B. R.; Huang, J.; Alyea, F. N.; Cunnold, D. M.,; Fraser, P. J.; Hartley, D. E.; Simmonds, P. G. Science 1995, 29, 187.
(13) Montzka, S. A.; Bulter, J. H.; Myers, R. C.; Thompson, T. M.; Swanson, T. H.; Clarke, A. D.; Lock, L. T.; Elkins, J. W. Science 1996, 272, 1,318. (14) Cunnold, D. M.; Weiss, R. F.; Prinn, R. G.; Hartley, D.; Simmonds, P. G.; Fraser, P. J.; Miller, B.; Alyea, F. N.; Porter, L. J. Geophys. Res. 1997, 102, 1,259. (15) Derwent, R. G.; Simmonds, P. G.; O’Doherty, S.; Ryall, D. B. Atmos. Environ. 1998, 32, 3689. (16) AFEAS. Production, sales and atmospheric release of fluorocarbons through 1994; AFEAS Science and Policy Services Inc.: Washington, DC, 1996. (17) Wang, J. L.; Chang, C. J.; Lin, Y. H. Chemosphere 1998, 36, 2391. (18) Koppmann, R.; Johnen, F. J.; Plass-Dulmer, C.; Rudolph, J. J. Geophys. Res. 1993, 98, 20517.
(19) Montzka, S. A. Climate monitoring and diagnostics laboratory, summary report 1994-1995; NOAA Environmental Research Laboratories: Boulder, CO, 1996; pp 23, 84. (20) Wingenter, O. W.; Kubo, M. K.; Blake, N. J.; Smith, T. W., Jr.; Blake, D. R.; Rowland, F. Sherwood J. Geophys. Res. 1996, 101, 4331-4340. (21) Wingenter, O. W.; Blake, D. R.; Blake, N. J.; Sive, B. C.; Atlas, E.; Flocke, F.; Rowland, F. S. J. Geophys. Res. 1999, 104, 21,81921,828.
Received for review November 21, 2000. Revised manuscript received April 16, 2001. Accepted May 21, 2001. ES001894Q
VOL. 35, NO. 16, 2001 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
3279
