Real-Time Fab-Wise Airborne Molecular Contaminant ( AMC ) Monitoring System Using Multiple Fourier Transform Infrared ( FTIR ) Spectrometers in a Semiconductor Plant

The objective of this study was to investigate the airborne pollutant emission sources and fluctuations around the indoor and outdoor environments of a semiconductor manufacturing plant using monitoring data that were collected over 4 consecutive days via three Fourier transform infrared (FTIR) spectrometers located near an outdoor make-up air unit and an indoor Fab and sub Fab. Based on a total of 1,032 five-minute-interval records, fourteen chemicals were detected. Six of these chemicals, namely, carbon tetrafluoride, nitrous oxide, carbon monoxide, silane, sulfur hexafluoride, and methane, had significant concentration correlations between the indoor and outdoor environments. With the exception of silane and sulfur hexafluoride, the percentage of indoor/outdoor concentration ratios that were greater than one ranged from 62.2% to 73.1%, indicating that the indoor chemical concentrations were typically higher than the outdoor concentrations. Based on the regression models derived for the indoor and outdoor nitrous oxide concentrations, the nitrous oxide was believed to be originally emitted from the sub Fab vented to the outdoors and then partially returned to the Fab. It was estimated that for one ppb of nitrous oxide detected in the Fab, 2.58 ppb of nitrous oxide could be detected at the make-up air unit, which might result from the sub Fab emission being at a high level of 6.60 ppb. Furthermore, elevated outdoor concentrations of chemicals, such as carbon tetrafluoride, nitrous oxide and carbon monoxide, were observed without previous indoor emission peaks, indicating that these chemicals might accumulate in the outdoor surrounding area. This study successfully illustrated the dynamically changing relationship between indoor and outdoor chemical concentrations in a semiconductor clean room. These results can be used to prevent subtle and potential adverse impacts of airborne molecular contaminant (AMC) in various manufacturing facilities of technological industries, including the semiconductor and optoelectronics industries.


INTRODUCTION
According to the guidelines published in the International Technology Roadmap for Semiconductors (ITRS) of 2011, the concentrations of AMC and particles should be minimized as the feature size shrinks to the nanoscale (AMC in five different categories result in distinct defects in an electronic device (Ayre et al., 2005)).For example, molecular acid (MA) causes corrosion problems (Higley and Joffe, 1996), molecular base (MB) induces T-topping damage (Ruede et al., 2001), molecular condensables (MC) change the deposited film property from Si-N to Si-O (Saga and Hattori., 1997) and hazes the lens of the photomask (Barzahgi et al., 2001), molecular dopant (MD) shifts the p-type and n-type properties (Lebens et al., 1996), and the particles fail the die when the size is greater than 1/3 of the printed gated length (Kitajima and Shiramizu., 1997).
In semiconductor manufacturing, many hazardous gases are used, such as pyrophoric silane (SiH 4 ) for thin-film deposition, corrosive chlorine (Cl 2 ) for dry etching, and toxic arsine (AsH 3 ) for ion implantation.Additionally, some perfluorocompounds (PFCs, e.g., SF 6 and NF 3 ) with high global-warming potential are applied as etchants or for cleaning the chamber.These raw materials and by-products are potential contaminants that result in product losses.For example, some of these chemicals may accidently be released from pipeline leakage or spill at the loading dock, which is not only harmful to employees but may also lead to shutdown of the manufacturing process (Muller et al., 2000).This impact is also highlighted by the ITRS (International Technology Roadmap for Semiconductors) yield enhancement committee in the annual publication for each technology node.AMC are potential yield killers for the manufacturing process; therefore, controlling AMC in emissions in the semiconductor industry is necessary to avoid potential losses in production yields (Li et al., 2005(Li et al., , 2007a(Li et al., , b, 2009)).
However, it is difficult to identify the episode profile in terms of emission sources and affected areas.To overcome this obstacle, open-path Fourier transform infrared (FTIR) spectrometers are an effective tool for environmental monitoring because of their features of continuous remote monitoring, rapid and multi-compound analyses, and path average measurements (Simonds et al., 1994;Li et al., 2007a, b;Lin et al., 2008) in comparison to the traditional time-consuming air sampling methods followed by chemical analyses in a laboratory (Chen et al., 2000;Lin et al., 2007Lin et al., , 2008)).Several studies have applied the FTIR technique to identify emission sources in semiconductor and optoelectronics manufacturing plants (Li et al., 2007a(Li et al., , b, 2009;;Tsao et al., 2011), pharmaceutical plants (Chang et al., 2009), water treatment facilities (Freddie et al., 2004), traffic road areas (Bradley et al., 2000), and industrial areas (Kagann et al., 1999;Walter et al., 1999).The chemicals that are detected by FTIR in semiconductor and optoelectronics plants include ammonia, nitrous oxide, ozone, silane, sulfur hexafluoride (SF 6 ), volatile organic compounds (VOCs), and perfluorocompounds; some of these chemicals are AMC in the semiconductor industry (Chien et al., 2007;Li et al., 2007aLi et al., , 2009;;Tsao et al., 2011).These studies successfully demonstrated the use of FTIR in environmental pollutant monitoring by identifying multiple exposure compounds and emission sources.However, most of these studies only applied a single FTIR spectrometer to monitor the different identified durations in continuous time periods.This design is weak, and it is difficult to overcome some temporal variation factors, such as wind speed, wind direction, and continuous or non-continuous emissions.
The indoor-to-outdoor (indoor/outdoor) ratio (I/O ratio) of pollutant chemicals is a useful measurement for exploring air quality in schools (Lee et al., 1997;Blondeau et al., 2005;Popupard et al., 2005) and residential areas (Hkukadia et al., 1998;Lidia et al., 2000;Wu et al., 2000;Niachou et al., 2008;Hsu et al., 2011) based on biological and chemical agents and particulate matter.Many of these studies focused on traffic emissions, coke production, weather changes or natural disasters, but few studies have focused on industrial emissions, particularly in the high-tech industry.Based on the continuous data collection feature of FTIR, a high number of indoor/outdoor ratios will be produced temporally to explore the emissions variation of the emission sources and to establish an equation to predict the indoor and outdoor concentrations.
The objective of the present study was to monitor the AMC in the indoor and outdoor environments of a semiconductor clean room using multiple FTIR systems in real time.The temporal ratios of the indoor and outdoor concentrations were used to explore the emission sources and their correlated concentrations to estimate the emissions occurrence and to prevent AMC issues.

Study Clean Room and Its Air-Conditioning System
This study was conducted in a 200-mm wafer foundry semiconductor manufacturing company located in the southeastern section of the Hsinchu Science Park in northern Taiwan.The study's four-floor Fab (Fig. 1(a)) was 84 meters wide by 84 meters long.The study site consisted of a dirty sub Fab (first floor), a sub Fab (second floor), a Fab (third floor) and a supply air plenum (fourth floor).The outdoor air (OA) was drawn into the Fab and processed by make-up air units (MAUs), which removed the dust via a high-efficiency particulate air filter (HEPA filter) and eliminated the acidic and basic gases via a spraywasher system.The air supply was composed of 20% fresh OA and 80% returned air and was drawn by 2,500 sets of fan filter units (FFUs) in the ceiling to supply unidirectional air flow.The air exchange flow of the make-up air was 115,200 CMH (m 3 /h), as supplied by six MAUs.

Fourier Transform Infrared (FTIR) Spectrometer
The FTIR instrument that was utilized in this study was manufactured by ETG (Air Sentry, Baltimore, Maryland, USA) and consisted of a Fourier transform spectrometer (FTIR system), a retroreflector, a gas cell, and a telescope.The FTIR system is composed of an IR source, an interferometer, a standard mid-band mercury cadmium telluride (MCT) detector, and an operational computer with RMM Soft software.An automatic liquid nitrogen filling system was equipped to maintain the MCT detector temperature at 77 K.In this study, two types of FTIR assemblies were used.The first assembly was an open-path FTIR that was used to detect any pollutants passing through the monitoring path between the FTIR and its retroreflector.The retroreflector consisted of a panel of gold-coated precision corner cubes that were mounted in a protective housing, and it was placed at the end of the measurement path facing the telescope.Two sets of this type of openpath FTIR (OP-FTIR) assembly were used in both the Fab and the sub Fab.An OP-FTIR can instantly determine the gas contaminants that are present in the monitoring path by detecting the infrared absorbance spectral fingerprint of each gas and using advancing computer calculations.To accurately operate the OP-FTIR and to analyze the collected infrared spectra, the US-EPA published guidelines and related literature were followed (US EPA; Malachowski et al., 1994;Farhat et al., 2000).The second assembly was a closed-cell FTIR that was used to draw the sampling gas into the gas cell for analysis.The gas cell is a one-meter tube in which the consecutive reflection lighting was designed to enlarge as a 40-meter optical path length to enhance its detection sensitivity.The closed-cell FTIR was used to monitor the pollutants at the MAU.

Sampling Location
As mentioned above, a total of three FTIR assemblies were simultaneously applied in this study to monitor the air quality at the MAU, the Fab, and the sub Fab.A closedcell FTIR was positioned outside an MAU to monitor the quality of the outdoor air that was drawn in through the MAU.For the Fab with an open-path FTIR, an 82-m-long monitoring path was set at the central walking path (Fig. 1(b)); the height of this monitoring path was greater than 1.8 m to prevent interference from workers on the spectrum as they passes through this FTIR monitoring path.The monitoring path of the open-path FTIR for the sub Fab was placed in front of the filter of the recirculation duct with a path length of 76 m to monitor the pollutants in the return air.

Using the Indoor/Outdoor Concentration Ratio to Identify the Emission Source
The indoor pollutant concentrations were monitored at the Fab and the sub Fab, and the outdoor concentrations were monitored at the MAU.The data were collected in 5-min intervals.For the samples with detectable chemical levels, the indoor/outdoor concentration ratio was obtained by dividing the outdoor concentration by the indoor concentration of the specified chemical matched by the sampling time period.

Data Collection and Analysis
During the monitoring period between 11 and 14 April, a power failure occurred in the Fab and resulted in a 2-hour monitoring loss on the afternoon of April 14, accounting for approximately 2.4% of the total 112-hour monitoring period.
The FTIR monitoring data were analyzed using a multilevel classical least-squares (CLS) method in combination with reference IR spectra that were generated from the U.S. EPA and the Infrared Analysis company databases.The modulated signal from the interferometer was an interferogram, which was Fourier transformed to produce a spectrum from 700 to 5,000 cm -1 at a resolution of 1 cm -1 .For each spectrum, 128 interferograms were accumulated and integrated, and an average of one spectrum was produced every 5 min.One of selected spectra of each monitoring line was identified for being the referent background to correct all detected spectrum.The referent background meant that only general atmosphere constituents are detected.This process should be done by experience analyzer to calibrate the data accurately.An infrared spectrum represents a fingerprint of a sample with absorption peaks that correspond to the frequencies of the vibrations between the bonds of the atoms comprising the compound.Because each compound consists of a unique combination of atoms, the intensity of the peaks in the spectrum could be a direct indication of the amount of compound that was present.Furthermore, the integrated area of each peak could be calculated to determine the concentration of the identified chemical.However, water has the potential to impact the analysis of absorption spectra whenever applying the FTIR, particularly the OP-FTIR on rainy days.The H 2 O and CO 2 in the atmosphere will generally interfere with the infrared spectral wavelength regions of 700 to 5,000 nm, except for 700 to 1300, 2050 to 2250 and 2850 to 3050 nm, indicating that these three spectral regions are the effective areas for OP-FTIR to identify chemicals.Moreover, compounds that were composed of two atoms, such as oxide, nitrogen, and chorine, could not be identified by FTIR because the vibration of the single bond could not be detected by FTIR.Determining the detection limits of the compounds analyzed with FTIR spectrometers depends on the infrared absorption coefficients of each compound, the optical path length, and the sensitivity of the instrument.The measurements may vary considerably if the concentrations of the target gas and the type of interfering chemicals change.In addition, the FTIR measurement reported in this study were the path average concentrations, which might occasionally underestimate the actual concentrations of the monitored samples given that the FTIR path length is greater than the width of the emission plume.This condition typically occurred during the early stage of peak detection before the emitted chemicals occupied the indoor environment.In this study, a detectable sample was defined as any sample measured over the detection limits, and the detected rate was calculated as the proportion of detectable samples in the total samples.
In this study, Pearson correlation analyses were used to explore the relationships between the indoor and outdoor concentrations.A simple linear regression was also used to examine the relationship of the indoor concentration with the outdoor concentration, with the indoor concentration as the independent variable and the outdoor concentration as the dependent variable.To avoid confounding outliers, the model analysis only included the matched indoor/outdoor samples with a concentration ratio between the 25 th and 75 th percentiles.The statistical analyses were performed with the Statistical Analysis System (SAS, version 9.1, Cary, North Carolina, USA).

Monitoring for Chemicals with FTIR
During the monitoring period, 1,032 records were collected by each FTIR.The fourteen compounds that were detected included ammonia, carbon tetrafluoride, nitrous oxide, carbon monoxide, perfluoroethane, silane, methane, sulfur hexafluoride, isopropanol, perfluorohexane, FC 3283 (perfluorocompounds, primary compounds with 9 carbons), 1,1,1,2-tetrafluoroethane, ethanol, and trifluoromethane.Among these identified compounds, 8 chemicals were detected at the MAU (make-up air unit), 13 chemicals were detected in the Fab, and 12 chemicals were detected in the sub Fab (Table 1).For all of the samples, none of the study chemicals exceeded the corresponding time-averaged threshold limit values (TLVs) (ACGIH, 2013).In addition, no odorous episode occurred during this monitoring period because no chemical was detected over the odor threshold value.
Table 1 shows the six chemicals that were only detected in the indoor environment during the preliminary emission source identification, suggesting that these chemicals were emitted inside the indoor environment (i.e., in the Fab and sub Fab).Among these six chemicals, isopropanol and ethanol were the major components of the common clean solvent that was used in the periodical maintenance of tools (Chien et al., 2003;Wu et al., 2003;Li et al., 2005).Perfluorohexane is the clean gas that was used in the etching process (Xiao, 2000).FC3283 is a cooler cryogen that is used to maintain the chamber temperature.Tetrafluoroethane and trifluoromethane are the byproducts of FC3283 in the temperature stabilization process.In contrast, ammonia was only detected at the MAU, indicating that it was emitted from the outdoor source, which is consistent with the findings of our previous study (Tsao et al., 2011).In addition to the above-mentioned chemicals, the other chemicals were simultaneously detected in both the indoor and outdoor environments.Temporal Profiles for the Detected Chemicals Fig. 2 presents the time-series plots for the fluctuations in the chemicals that were detected in both the indoor and outdoor environments.The temporally separated peaks of methane (Fig. 2 most likely were independent episodes.Following the occurrence of certain chemical peaks that were detected in the indoor environment, the consequential elevation in the concentrations of these chemicals in the outdoor environment suggested that the concentration fluctuations of these outdoor chemicals were influenced by indoor emission episodes and vice versa.The peak emission episodes for silane, methane, and sulfur hexafluoride resulted from indoor emissions and lasted approximately 45 min, 50 min and 90 min, respectively.These long time periods may provide the potential for airborne molecular contaminant (AMC) issues that might affect the semiconductor manufacturing process by damaging production yields.
Carbon tetrafluoride, nitrous oxide, and carbon monoxide were commonly detected in both the indoor and outdoor environments.Among the peaks of these monitored chemicals, the concentrations of nitrous oxide as measured in the Fab were generally higher than those in the sub Fab, which were higher than those in the MAU.By combining the plots of carbon tetrafluoride (Fig. 3) and nitrous oxide, we identified a similar temporal profile between the indoor and outdoor environments, particularly in the first three monitoring days, indicating that both carbon tetrafluoride and nitrous oxide might originally originate from the same emission source.Although carbon monoxide also presented a profile that was similar to that of carbon tetrafluoride and nitrous oxide, the outdoor carbon monoxide concentrations were not as similar as those of the data in the Fab and sub Fab.

Correlation between the Indoor and Outdoor Chemical Concentrations
According to the detected rate (DR) for the chemicals at the MAU sampling site (Table 1), carbon tetrafluoride, nitrous oxide, and carbon monoxide were the three most detected chemicals, with DRs of 29.6%, 12.7%, and 7.5%, respectively.In the Fab and sub Fab, which were considered to be indoor environments, the chemicals with a DR that was greater than 10% included nitrous oxide (48.0% in the Fab), carbon monoxide (40.9% in the sub Fab), isopropanol (39.0% in the Fab), carbon tetrafluoride (25.3% in the sub Fab), sulfur hexafluoride (25.3% in the Fab), ethanol (15.2% in the Fab), perfluoroethane (12.1% in the sub Fab), FC 3283 (10.6% in the Fab), and methane (10.1% in the sub Fab).Based on the synchronous and continuous FTIR monitoring, the time-matched samples that were detected at the MAU and in the Fab or the sub Fab were included in the correlation analyses.Table 2 shows a significant correlation between the chemical concentrations that were detected in the Fab and at the MAU for nitrous oxide, silane, and sulfur hexafluoride (p < 0.05).The same results were also observed for the concentrations of carbon tetrafluoride, nitrous oxide, carbon monoxide, and methane between the sub Fab and the MAU.
As shown in Table 2, nitrous oxide was the only chemical that presented a significant concentration correlation between the Fab and the MAU and between the sub Fab and the MAU.In addition to nitrous oxide, the concentration correlations for carbon tetrafluoride and methane were statistically significant between the MAU and the sub Fab.In addition, the concentration correlations for silane and sulfur hexafluoride were also statistically significant between the MAU and the Fab.

Indoor/Outdoor Concentration Ratio for the Identification of the Emission Source
The results of the indoor/outdoor concentration ratio (I/O ratio) for the correlative chemicals are summarized in Table 3.Based on the definition of the indoor/outdoor concentration ratio, a mean ratio that was greater than one signified a higher indoor concentration.With respect to the percentile distribution of the indoor/outdoor concentration ratio, the median indoor/outdoor concentration ratios for the study chemicals in Table 3 were generally greater than one, except for silane and sulfur hexafluoride.Additionally, carbon tetrafluoride, nitrous oxide, and carbon monoxide were hypothesized to be primarily emitted indoors because the percentages of the indoor/outdoor ratios that were greater than one were 62.2% for carbon tetrafluoride in the sub Fab, 71.8% for nitrous oxide in the Fab, 73.1% for nitrous oxide in the sub Fab, and 64.0% for carbon monoxide in the sub Fab.The identified emission sources temporarily changed between the indoor and outdoor environments.The concentration peak of carbon tetrafluoride as shown in Fig. 2 indicated a higher indoor concentration, suggesting that the source might be located indoors.In contrast, in some cases in which the I/O ratio was less than one, the outdoor concentrations might be greater but without distinct previous concentration peaks as detected indoors.Based on this situation, it was generally believed that some of the carbon tetrafluoride that was emitted indoors might accumulate outdoors around the manufacturing site.The same interpretation may also hold true for the profiles of nitrous oxide and carbon monoxide.The identification of the sources of carbon tetrafluoride emissions was also supported by the etching process, which uses carbon tetrafluoride as a raw material for etching gas in the manufacturing process (Xiao, 2003).These gases have been categorized as greenhouse gases (GHGs) and perfluorocarbons (PFCs), whose reduction was an important mission that was declared by the RIO Earth Summit '92 and Kyoto Protocol-97 (Li et al., 2000a, b).One characteristic of this gas is the difficulty of eliminating it through natural reactions, resulting in a side effect of increasing atmospheric temperatures, also known as global warming.On the other hand, both nitrous oxide and carbon monoxide are raw materials that are used in the thin-film process for layer growing (Xiao, 2000).The observation of their existence in the environmental measurements in the present study might not be confounded by traffic exhaust because the MAU sampling site was set at over 21.5 m in height, which is sufficiently high to avoid any interference resulting from traffic emissions.
These aforementioned results and discussions supported the existence of temporary dynamic changes in the indoor and the outdoor chemical concentrations, which are highly dependent on whether the chemical was emitted from the indoor manufacturing process during the observation period and the wind speed in the outdoor environment (Tsai et al., 2004;Chiu et al., 2005).The chemicals that accumulate around the outdoor environment of a manufacturing facility may be drawn into the facility again and consequently deteriorate the indoor environment (Li et al., 2009).

Concentration Association between Indoor and Outdoor Chemicals
A simple linear regression model was used to determine the relationship between the indoor and outdoor chemical concentrations (Table 3).For commonly detected chemicals, such as carbon tetrafluoride, nitrous oxide, and carbon monoxide, Equation III shows high R 2 levels ranging from 0.46 to 0.89 based on samples with indoor/outdoor concentration ratios ranging between the 25 th and 75 th percentiles compared to Equation I and Equation II, which were, respectively, based on all of the detectable samples and samples with indoor/outdoor concentration ratios ranging between the 5 th and 95 th percentiles.In these models with the indoor concentration as an independent variable and the outdoor concentration as a dependent variable, a negative intercept (α) indicates a higher background chemical concentration as detected in the indoor environment compared to the outdoor environment.However, all of the regression coefficients (β) in Equation III for the six chemicals that are shown in Table 3 were between zero and one, indicating that the outdoor chemical concentrations were positively correlated with the indoor ones and that the magnitude of the chemical concentration increment was greater in the indoor environment than in the outdoor environment.Using carbon tetrafluoride in the sub Fab as an example, the indoor background chemical concentration was greater than the outdoor concentration by 0.25 ppb (α = -0.25).The β value of 0.88 indicated that the outdoor chemical concentration increased by 0.88 ppb, given that  the indoor chemical concentration increased by one ppb.However, Equation III for nitrous oxide in the Fab was an exception because the resulting α value was positive, 2.18 ppb, indicating that the outdoor background nitrous oxide concentration was 2.18 ppb higher than the indoor concentration.These differences were also found in the outdoor concentrations, varying from 2.18 ppb to 3.63 ppb (Fig. 3).From the turning-point concentration of 3.63 ppb, the outdoor nitrous oxide concentration was surpassed by the indoor nitrous oxide concentration.The concentration of nitrous oxide in the sub Fab could be obtained based on the nitrous oxide concentration in the MAU with regression Equation III specifically derived for nitrous oxide in Table 3.Following the regressions for nitrous oxide concentrations between the MAU and the sub Fab and between the MAU and the Fab, the concentration of nitrous oxide at the MAU would be 2.58 ppb given that the indoor Fab nitrous oxide concentration is 1 ppb, and the nitrous oxide concentration in the sub Fab would be 6.60 ppb.The resulting relationship of nitrous oxide concentrations among the Fab, sub Fab, and MAU could be used to determine the emission source in the sub Fab when presented with the maximum nitrous oxide concentration among these three sampling locations.However, if the nitrous oxide concentration at the Fab exceeded the turning-point of 3.63 ppb, for example, the concentration in the MAU would be estimated to be 3.78 ppb given the Fab concentration of 4.00 ppb.Therefore, the sub Fab concentration could be calculated as 8.60 ppb with another regression model of the MAU and sub Fab.
Based on the nitrous oxide concentrations in these three locations, we could also conclude that the nitrous oxide emission was originally from the sub Fab.This situation could be explained by the fact that the recirculated air in the Fab was composed of 80% returned air from the sub Fab and 20% fresh air from the MAU.The concentrations of nitrous oxide in the Fab and sub Fab were, therefore, highly correlated, with a statistically significant correlation coefficient of 0.58 (p < 0.0001).In contrast, with relatively low levels of nitrous oxide and accounting for the low percentage of the circulated air in the Fab, the outdoor nitrous oxide originating from the MAU appeared to be diluted in the recirculated-air flow in the Fab.In general, the nitrous oxide that was emitted from the sub Fab directly dominated the Fab nitrous oxide concentration through the role of a major chemical component in the recirculated air.

Airborne Molecular Contaminant (AMC) Impacts on the Fab Operation
According to the prior analysis and discussion, the notorious yield killer, i.e., ammonia, is successfully controlled by the MAU, but the detectable period was as long as 4.3 hours, which potentially affects the Fab operation when the MAU malfunctions or has a low efficiency.On the other hand, several preventive maintenance (PM) compounds, such as isopropanol and ethanol, were generally detected in the Fab.Without well-trained engineering skill and a functional local exhaust system when operating the PM, the high concentrations and detectable rate will potentially increase the explosion risk.Finally, the greenhouse gases (GHGs) group, e.g., carbon tetrafluoride, nitrous oxide and carbon monoxide, was typically found indoors and outdoors because their characteristic was hardly reacted or eliminated.However, as high energy power technology is applied for new processes, such as E-beam to replace EUV (extreme ultraviolet) in the lithograph process, more than ten times the amount of energy will be applied (Bozano et al., 2011).It is possible that future studies should focus on whether these GHGs compounds or their by-products could react to become a potential impact.

CONCLUSIONS
With synchronous monitoring using three FTIR spectrometers, this study successfully illustrated the dynamically changing relationship between the indoor and outdoor chemical concentrations in a semiconductor clean room.The major emission source was identified and most likely originated from inside the facility based on the elevated outdoor chemical concentrations following the increased indoor concentrations.Elevated concentrations of outdoor chemicals, such as carbon tetrafluoride, nitrous oxide, and carbon monoxide, were observed without previous indoor emission peaks, implying that these chemicals might accumulate in the outdoor surrounding area.This study demonstrated that multiple FTIR measurements are effective in helping explore pollutant emission sources and transportation pathways for a manufacturing facility of emerging high technology, such as the semiconductor industry and optoelectronics industry.

*Fig. 1 .
Fig. 1.The study clean room, the air flow directions and the locations of the FTIR setup.
Fig. 2. Time-series plots of the path average concentrations for seven study chemicals during the 112-hour observation period.

Fig. 3 .
Fig. 3. Regression for indoor nitrous oxide concentrations with the outdoor nitrous oxide level.

Table 1 .
Monitored concentrations and detectable rates of chemicals by sampling site.
d DL is the method detection limits for each FTIR in this study.e Odor threshold values cited from the National Institutes of Health, US. f Not available.

Table 2 .
Correlation between outdoor and indoor chemical concentrations.the number of detectable samples; b the number of time-matched detectable samples, as being matched to the samples at the MAU. a

Table 3 .
Indoor/outdoor concentration ratio by chemical and simple linear regression model for the relationship of outdoor chemical concentration with indoor concentration.