Characteristics and Source Apportionment of Ambient Volatile Organic Compounds in a Science Park in Central Taiwan

Air samples were collected concurrently at four sites using stainless steel canisters in a science park in central Taiwan. The airborne volatile organic compounds (VOCs) were then analyzed using a gas chromatograph and a mass spectrometer (GC/MS). Eighteen volatile organic compounds (C1–C8) in six hydrocarbon groups were identified. Measurements reveal that the three dominant species were 2-butanone (8.60 ± 0.98 ppbv), toluene (6.13 ± 1.67 ppbv), and acetone (6.03 ± 2.79 ppbv), while most other species were present at a concentration of below 1.00 ppbv. On average, the most abundant hydrocarbon group was ketones (56.95%), followed by aromatic compounds (27.75%), alkanes (8.33%), fluoroalkanes (3.40%), chloroalkanes (2.47%), and nitrile compounds (1.10%). Principal component analysis (PCA) identified two components PC1 and PC2. Ten species in PC1 and eight species in PC2 had loadings of greater than 0.8, suggesting that the emission sources of PC1 were related to high-tech industries and traffic, and those of PC2 were related to fugitive emissions of organic solvents and refrigerants.


INTRODUCTION
High-tech industries, such as the integrated circuit industry, semiconductor industry, optoelectronic industry and biotech industry, are currently the main economic activities.Along with the growth of high-tech industries, they bring us all kinds of chemicals, including silanes, silicon chemicals, halogenated chemicals, inorganic acids, caustics, and volatile organic compounds (VOCs).Similar situations can be found in other industrial and developing countries (Su et al., 2006;Cai et al., 2010;Leuchner and Rappengluck 2010;Tiwari et al., 2010).
Many works have also demonstrated that oil refineries, chemical and plastic factories and semiconductor manufacturers are major sources of VOCs in the atmosphere (Nikolaou et al., 2002;Khwaja and Narang, 2008;Nian et al., 2008;Cai et al., 2010;Leuchner and Rappengluck, 2010;Yuan et al., 2010;Vega et al., 2011).In addition, many VOC species are also precursors of ground level ozone formation and they are mainly contributed from on-road mobiles and industrial factories as to worse the air quality and reduce atmosphere visibility (Zhang et al., 2009;Vardoulakis et al., 2011;Lin et al., 2012;Peng et al., 2013).
Studies have shown that toluene to benzene (T/B) ratio of approximately 2 is a good predictor of traffic exhaust (Tang et al., 2008;Civan et al., 2011;Wang et al., 2012); meanwhile, the T/B ratio increases with the emission strength of industrial source (Peng et al., 2013).
US EPA (1990) showed that ambient VOCs attribute to 35-55% of the outdoor air cancer risks in the United States.According to the "Hazardous Air Pollutants (HAPs) List", published by US-EPA (1996), approximately 70% of 189 HAPs are VOC species.The emissions of these species expose workers or nearby residents to risks associated with mutagens and carcinogens (Liao et al., 2004;Ladou and Bailar, 2007) and should be regulated and controlled carefully.
As part of the implementation of the "Silicon-Island Policy", Taiwan's government has established many science parks in Taiwan.However, the impact of the emissions of VOCs, acidic and alkaline gases from the high-technology factories in the environment and human health is a matter of serious public concern.Monitoring programs and emission controls for VOCs and HAPs at the science park are important, but few relevant measurements have been made.
This work is part of an integrated research that comprises three sub-projects to study the emissions of HAPs from the science park in central Taiwan, and their associated health risks.This paper identifies the VOC species and their concentrations in the atmosphere around the science park.Principal component analysis (PCA) method is conducted to identify potential sources of the detected species.

Sampling and Analysis
The science park studied herein is in Houli in Taichung City in central Taiwan (latitude 24.317-24.323 'N, longitude 120.718-120.Airborne VOCs were collected concurrently at four sites S1, S2, S3, and S4 in the science park on March 22 in 2012 (Fig. 1).Site S1 was 960 m northwest of Plant A; S2 was 680 m northeast of Plant A; S3 was 180 m southeast of Plant A, and S4 was 50 m southwest of Plant A. The weather was fairly well during the sampling period.Air temperature was 21.5-22.4°C,pressure was about 1002.8 hPa, and relative humidity was 57-62%.A southerly wind prevailed at speed of 0.9-1.4m s -1 .
In compliance with the procedures of Taiwan-EPA NIEA-A715.14B,which are equivalent to those of the Compendium Method TO-15 (US-EPA, 1999), air samples were collected for 2 hr at a flow rate of 40 mL min -1 using 6 L stainless steel canisters (Silonite Summa Style, EN TECH, USA).Before sampling, all canisters were cleaned, moisturized, and checked for leaks to ensure a vacuum pressure of less than 10 -2 mm Hg.The air samples were then analyzed using a gas chromatograph (GC, Agilent 6890N) and a mass spectrometer (MS, Agilent 5973N).The GC oven temperature was set to 40°C initially, rising to 50°C after 2 min, and then increasing at 8 °C min -1 to 230°C, which temperature was held for 10 min.
Six-point calibrations were performed for each species, yielding a linear regression with a coefficient of determination, R 2 , above 0.995.Due to limited spaces, calibration curves or control charts for various VOC species were not presented.The detection limit (DL) for each species was implemented following US-EPA Method TO-15 65 component mix (RESTEK cat. No. 34436).Table 1 shows that the DLs for 18 species identified in this work ranged from 0.17 to 0.38 ppbv, with recovery efficiencies in 96.2-103.9%.

Principal Component Analysis
Source apportionment of VOC species was performed using principal component analysis (PCA), based on the varimax orthogonal rotation method.Data entry and analysis were done using SPSS statistical package version 17.The main purpose of the PCA is to describe the covariance relationships among many variables in terms of just a few underlying common components, which hopefully explain the most (typically 80 to 90%) of the total variance in the observed data (Johnson and Wichern, 2007).A component that is identified from PCA, typically with an eigenvalue of greater than one, exhibits a pattern of variation in response to the input parameters (which are the concentrations of pollutants herein).The correlation between the concentration of a particular pollutant and a component increases with its A: semiconductor company; B: optoelectronics company.(Derwent et al., 1995;Ho et al., 2002).

Characteristics of Atmospheric VOCs around Science Park
Fig. 2 plots the concentrations of 18 VOC species at four sampling sites.Site S1 exhibited the highest concentrations of 2-butanone (10.30ppbv) and toluene (8.20 ppbv), while site S2 had the highest concentration of acetone (8.50 ppbv).The concentration profiles at the four sites were clearly similar in that 2-butanone (8.60 ± 0.98 ppbv), toluene (6.13 ± 1.67 ppbv), and acetone (6.03 ± 2.79 ppbv) were the three dominant species, and the other species were minor with concentrations of less than 1.00 ppbv.These three species are commonly utilized as solvents or cleaning liquids in semiconductor plants (Pethrick and Rankin, 1998;Chan and Lai, 2010;Tsai et al., 2011).
The contours of TVOCs with the corresponding wind rose during the survey period is plotted in Fig. 3. TVOC was the highest at S1 (35.09 ppbv), followed by S2 (30.03 ppbv), S4 (29.23 ppbv), and S3 (14.32 ppbv).Since a southerly wind prevailed with a speed of 0.9-1.4m s -1 , the downwind (northeast) sites S1 and S2 exhibited higher TVOCs than the upwind (south or southwest) sites S3 and S4.

Ratio of Toluene to Benzene (T/B Ratio)
Toluene and benzene are organic solvents commonly used in semiconductor processes (Park et al., 2011;Park and Yeo, 2013), but they are also present in vehicle emissions.The percentages of benzene, toluene and hexane in TVOC at S1, S2, S3, and S4 ranged in 0.00-1.42%,19.64-25.14%,and 0.00-1.45%,respectively, indicating that toluene was pre-dominant in this work.Table 3 compares the T/B ratios found in various studies, which were 14.8-18.9here, consistent with other works.For example, the T/B ratio was 14.3 at the Lin-Yuan Industrial Park in Kaohsiung, 14.3 at Guangzhou (industrial) in China, and 11.3 at Bursa industrial city in Turkey.Civan et al. (2011) found that T/B > 10 is typically associated with an industrialized city and local sources.Table 3 also reveals that the T/B ratio was 5.2 at Tsuen Wan in Hong Kong and 7.0 at Urayasu city in Japan.Guo et al. (2004) and Civan et al. (2011) found that T/B < 10 is typically associated with a non-industrialized city where the main pollution source is traffic exhaust.Table 3 reveals that the measured T/B ratios were 2.5 to 3.5 on various road-sides or highways, where vehicle exhausts were the primary sources.

Source Apportionment Using PCA
Table 4 shows loadings of 18 VOC species categorized in two components obtained from PCA.The first component PC1 is the main one, which explained the 67.37% of total variance, while the second component PC2 explained the remaining (32.63%); that is, these two factors together explained all the variances (100%).
The PCA results discussed above suggested that the possible primary sources of PC1 were related to high-techs (e.g., semiconductor and optical-electrical industries), while the possible secondary sources of PC2 were related to organic solvents and refrigerators (see Table 4).Table 5 further summarized the possible pollution sources of the18 VOC species from other studies, which included semiconductor, electronic and machinery factories, vehicles, gas stations, organic solvents, and refrigerants.Comparisons between Tables 4 and 5 indicate that the possible pollution sources suggested here were reasonable.
Source apportionment using PCA identified two source components.The primary component PC1 explained the 69.77% of total variance, while the second component PC2 explained the rest (30.23%).Ten species in PC1 and eight species in PC2 had loadings of more than 0.8.These results suggest that the emission sources of PC1 were associated with the high-tech industries (e.g., semiconductor and optical-electronics) and vehicle exhausts; and those of PC2 were associated with fugitive emissions of organic solvents and refrigerants.
737 'E) near Feng-yuan city; it has an area of 134.4 hectares.The third and fourth phases of its development began in 2008 to attract semiconductor, optical-electrical and fine machinery factories.During the study period in March 2012, only two plants A and B were successfully operating while others were still under construction.Plant A mainly produced DRAM (Dynamic Random Access Memory), and plant B mainly produced µc-Si tandem solar PV modules.The allowable emissions for VOCs issued by the Environmental Protection Bureau of Taichung City were 18.10 ton per year for Plant A and were 2.58 ton per year for plant B. According to the emission inventory of TEDs-8.1 (2014), the emissions of NMHCs in Houli were about 17 tons per year, and those of traffic sources were about 60.6 tons per year.The concentration of NMHCs at the Houli air-quality station operated by Taiwan-EPA was 450 ppb during the sampling period, indicating that the influence of ambient concentration was primarily due to local traffics.

Fig. 1 .
Fig. 1.Four sampling sites in the science park.

Table 1 .
Results of quality assurance and quality control of 18 VOC species identified.

Table 2 .
Average concentrations (ppbv) of 18 VOC species at four sites.

Table 3 .
Comparisons of toluene to benzene ratios (T/B).

Table 4 .
Factor loadings obtained from PCA.

Table 5 .
Potential emission sources of VOCs taken from the literature.