FTIR spectrometers measure scrubber abatement efficiencies
07/01/2002
Shou-Nan Li, Jung-Nan Hsu, Hui-Ya Shih, ITRI
Shu-Jen Lin, Jeng-Liang Hong, Winbond Electronics Corp., Hsinchu, Taiwan
Overview
The extremely toxic, corrosive, and pyrophoric nature of process exhaust gases poses a major challenge to a semiconductor manufacturing company. To tackle the problem, Winbond Electronics Corp. began a company-wide abatement efficiency evaluation of local scrubbers (or point-of-use abatement instruments) and central scrubbers using two extractive FTIR spectrometers. These activities resulted in a cost reduction estimated by Winbond to be greater than US$1 million annually because of the increased cartridge lifetime for dry scrubbers. Insurance premiums were also reduced as a result of the lower risk associated with process gases.
Members of the World Semiconductor Council, including the US, Europe, Japan, Korea and Taiwan, have voluntarily committed to reduce perfluorocompound (PFC) emissions substantially because of their potential global-warming effects. Mitigating the impact of process exhaust gases is achieved in a number of ways. Besides the usage optimization (or minimization) of process gases [1], local scrubbers (also called point-of-use abatement instruments) are installed right after the process tools to treat exhaust gases. Central scrubbers are used (usually at the fab roof) to handle acid/base gases or volatile organic compounds (VOCs) for general exhaust from ventilation hoods and from local scrubbers.
In the semiconductor industry, scrubbers used for treating exhaust gases are typically classified into three categories: 1) thermal oxidization, 2) dry adsorption and 3) wet scrubbing. General application of these types of scrubbers is described by Hayes and Woods [2], but in situ measurement results (other than manufacturers' advertisements) that would help in choosing commercially available scrubbers are limited. The study results help identify the hazards of process exhaust gases, constitute the database of scrubber performance for future purchasing reference, reduce the scrubber running costs, and improve abatement efficiency to decrease the exhaust gas risks inside the fab and pollutant emissions to the environment.
Local and central scrubbers widely used for treating exhaust gases were evaluated at Winbond. The local scrubbers evaluated include 13 thermal types (six DAS ESCAPE, six CDO-858, one TPU), four dry adsorption systems (three EBARA, one CLEAN-TECH) and four wet scrubbers (three ECOSYS VECTOR, one KPC). With respect to central scrubbers, nine wet scrubbers and two VOC rotating drums were evaluated.
 |
Material and methods
Information about the local scrubbers evaluated is listed in Table 1, including the energy and water consumption rates. Manufacturing processes that were a part of the study include dry etching, ion implantation, thermal diffusion and chemical vapor deposition (CVD). A central (acid/basic) wet scrubber, similar to the local wet scrubber, applies facility water to remove water-soluble gases. Plastic packing materials increase water-gas contacting surfaces. On the average (based on 10 samples), the water consumption rate for each central scrubber is about 9 liter/min (lpm).
A (central) rotating drum uses adsorption materials (e.g., zeolite) to remove VOCs from the gas stream. To make the rotating drum continuously function, part of the adsorption materials are kept regenerated by some (~7%) of the gas stream heated to a temperature of about 180°C. This heated gas stream passes through the spent adsorption materials and desorbs the VOCs from the rotating drum. Then, the desorbed VOCs are abated using a thermal oxidizer. Generally, natural gas is used as a fuel for the thermal oxidizer at a consumption rate of about 700 lpm.
Sampling set-up
Destruction removal efficiency (DRE) of a (local) scrubber is calculated based on inlet (i) and outlet (o) mass flow rates as shown:
DRE = 1-(Mo/Mi) (1)
If Ti = To and Pi = Po
DRE = 1 -(Co/Ci) (Qo/Qi (2)
where:
M = gas mass flow rate, T = gas temperature, P = gas pressure,
C = gas volumetric concentration, Q = gas volumetric flow rate.
The DRE calculation can be simplified to Eqn. 2 if Ti = To and Pi = Po. Using Eqn. 2, the DRE value is calculated using the gas concentration ratio (Co/Ci) times the gas flow dilution ratio (Qo/Qi).
 Figure 1. Schematic diagram of the sampling set-up for local scrubber evaluation. |
During this study, no FTIR measurement guideline in Taiwan existed, so the US Environmental Protection Agency (EPA) Method 320 protocol [3] was used. To prevent airborne particles from contaminating the FTIR gas cell, a Teflon membrane filter with a pore size of 0.8mm was enclosed inside a stainless-steel holder and connected to the inlet of the FTIR sampling train. For in situ local scrubber evaluation (Fig. 1), Ci entering the local scrubber was obtained using an extractive Fourier transform infrared (FTIR) spectrometer located at the scrubber inlet. Co was determined using a second extractive FTIR located after the local scrubber. Because of the large gas concentration difference before and after the local scrubber, the front FTIR had a gas cell with an optical path length of 0.01 or 0.1m, while the back FTIR had a gas cell with an optical path length of 1 or 10m.
It was found that both Ti and To were close to 22°C, and were measured using two thermocouples separately connected to the sampling pipelines right after the (front and rear) FTIR gas cells. Using two pressure gauges separately connected to the FTIR gas cells, both Pi and Po were found to be close to 750mm-Hg. Because of the addition of high-flow rate (10-50 lpm) house nitrogen gas to the pump located between the process tool and the local scrubber, Ti is close to 22°C and Pi is close to 750mm-Hg. This N2 addition would quickly change the exhaust gas temperature and pressure to be near that of the room.
Similarly, for To and Po, the addition of high-flow rate (100-500 lpm) compressed dry air (CDA) at the exit of the local scrubber also renders the treated exhaust gases to room conditions. In addition, for thermal-type local scrubbers (e.g., DAS-ESCAPE, TPU, CDO-858), the wet scrubbing system behind the thermal reaction chamber cools the hot exhaust gases such that they are close to room temperature. For central scrubber evaluation, since Qi = Qo, the DRE calculation is further simplified to be: DRE = 1-Co/Ci.
To determine Qi and Qo, three different techniques were used. Besides the SF6 injection method [4], direct flow rate measurement of every air or gas flow entering the local scrubber was conducted using a calibrated dry gas meter or a calibrated mass flow meter. Another flow rate measurement technique involved the introduction of inert pure PFCs (e.g., CF4, C2F6) of a known flow rate (~50 sccm) into the process chamber with the plasma or thermal mechanism turned off and the heating or combustion source of the local scrubber also turned off. The flow rate measurement technique chosen depended on the feasibility of the on-site conditions.
Treating CVD process exhaust
A DAS ESCAPE is commonly used to abate the process exhaust gases coming from a thin-film or a dry etching process. Based on in situ measurement results, this model has a DRE >99% for most exhaust gases except PFCs (Table 2). For CF4 (a very stable PFC), the DAS ESCAPE has a DRE value ranging from 80-91%. Generally, the abatement efficiency decreases as the inlet flow rate Qi increases. Typical scrubber inlet and outlet mass flow rates of CF4 are shown in Fig. 2a, which shows the effectiveness of the DAS ESCAPE for treating PFCs. In general, compared to the combustion-type scrubber, the abatement efficiency (or the DRE value) of the DAS ESCAPE compares well with that of the TPU discussed below.
 |
Although the DAS ESCAPE can effectively treat exhaust gases, undesirable by-products (CO, NO2, and CH4) are produced at the exit of the scrubber, as shown in Fig. 2b. The concentration and duration of these by-products are highly variable. In addition, for an HDP (high-density plasma) process chamber of a Centura-5200 (manufactured by Applied Materials) using SiH4 as a deposition gas and NF3 as a chamber cleaning gas, some CF4 gas was measured at the outlet of the DAS ESCAPE. As shown in Fig. 2c, the appearance of the exiting CF4 peaks matches well with that of the entering NF3 peaks. This means that, at an elevated temperature (>800°), CF4 gas could be produced from the reaction of methane (CH4) with F radicals or F2 gas (as described by Burgess et al.) [5]. To prevent the CF4 formation due to the combination of CH4 with F2, a wet scrubbing system could be added at the inlet of the local scrubber to remove F2 gas or replace CH4 with H2 as a combustion fuel.
 Figure 3. Inlet and outlet mass flow rate variation of a) CF4 and b) PH3 gas for a CDO-858. |
The PH3 hazard
Traditionally, a CDO-858 was installed to treat the exhaust gases from most processes. However, in situ measurement results found that a typical CDO-858 had good abatement efficiencies only for highly water-soluble gases (e.g., HF, SiF4) or for highly pyrophoric gases (e.g., SiH4). For PFCs, such as CF4 and NF3, CDO-858 has very limited abatement efficiency (Fig. 3a). On-site evaluation also found that a CDO-858 had an average DRE value of 70% for PH3. If not properly treated, PH3 could cause some safety and health problems because of its highly toxic and flammable nature.
To reduce the PH3 hazards, a follow-up test was conducted to improve the abatement performance of this CDO-858. As Fig. 3b and Table 3 show, its DRE values for treating PH3 were increased by decreasing the entering air and nitrogen flow rates. Theoretically, decreasing the entering flow rate would increase the exhaust gas temperature and retention time inside the reaction chamber, both of which would enhance the DRE value.
 |
When the N2 flow rate was 38 liter/min and the air flow rate equaled 96 lpm, the DRE value was only 65% (Test 1, Table 3). When the N2 flow rate was decreased to 10 lpm and the air flow rate to 52 lpm, the average DRE value was increased to 85%. In addition, for SiH4 gas, decreasing the N2 and the air flow rate also increased its DRE value from 93% to 97%. Because the N2 and air flow rates of 10 lpm and 52 lpm were close to the lowest flow rate settings of the CDO-858, no further airflow reduction tests were conducted.
CVD and furnace processes
A TPU can effectively treat exhaust gases from WSi-CVD processes [4]. In this study, two FTIRs and two residual gas analyzers (RGAs) were simultaneously used to assess the exhaust gas hazards during an emergency shutdown condition. A TEL-WSi-CVD tool was used in conjunction with the following process gases: WF6, SiH2Cl2, and ClF3.
Besides some unused process gases, by-products of HF, HCl, SiF4 and Cl2 were also identified. During an emergency shutdown, the TPU sucks the exhaust gases into its reaction chamber for 20 sec before the exhaust gases are by-passed to the pipeline. The on-site test results found that, during an emergency shutdown, the gas concentrations inside the TPU outlet and by-pass pipelines were both below the detection limits of the FTIRs and the RGAs.
 |
An EBARA dry scrubber is usually used to remove the toxic or hazardous gases emitted from furnace processes. As Table 4 shows, this model has excellent abatement efficiencies for removing these hazardous gases, but the running cost of this dry scrubber is much higher than that of other types of dry scrubbers. In an attempt to find a lower-cost alternative, a CLEAN-TECH dry scrubber was also evaluated. The in situ evaluation found that a CLEAN-TECH dry scrubber has DRE values > 99% for SiH4 and > 95% for AsH3 gas. The SiH4 and AsH3 gas concentrations at the outlet of the CLEAN-TECH dry scrubber were both below the detection limits of the FTIR, which are 0.1ppm for SiH4 and 0.2ppm for AsH3.
Catalytic adsorption materials
In addition to trying other local scrubbers in lieu of the EBARA dry scrubbers, we also conducted a test for lengthening the usage lifetime of the catalytic adsorption materials for treating ClF3 gas and its by-products (SiF4, HF, and HCl) with an extractive FTIR monitoring the outlet gas compositions. The results showed that a 68% lifetime increase was achieved with all incoming gases treated.
Etching
An ECOSYS VECTOR or a KPC wet scrubber is normally used to treat the exhaust gases from a dry etching process. The on-site test results show that the wet scrubber has good abatement performance (>99%) only for highly water-soluble or water-reactive gases, such as HCl, HF, SiF4, or COF2. However, for SiH4 and CO gases, the DRE values were both < 30%. For PFCs (e.g., NF3, CHF3, and SF6), the DRE values were all close to zero.
 Figure 4. SiH4 volumetric concentrations at the inlet and outlet of a central wet scrubber. |
Central wet scrubbers also have good abatement performance (>99%) but only for highly water-soluble or water-reactive gases. For SiH4, as shown in Fig. 4, the DRE value is < 10%. With respect to VOC treatment, central rotating drums can effectively treat large-molecule compounds, such as propylene-glycol-monomethyl-ether-acetate (PGMEA) or polyol. However, for methanol, the average DRE value is only 23% (based on three samples). Typical test results for methanol are shown in Fig. 5.
 Figure 5. CH3OH volumetric concentrations at the inlet and outlet of a VOC rotating drum. |
Conclusion
Scrubber performance and exhaust gas compositions were successfully acquired using FTIRs and RGAs. By reducing the entering N2 and air flow rates, the abatement efficiency of the CDO-858 for treating PH3 gas was improved from 65% to 85%. It was also found that the cartridge usage lifetime of the dry scrubber could be greatly increased by monitoring scrubber outlet gas concentrations. The database generated during this study was used to evaluate the fitness of the piping materials, to mitIgate the hazards of process exhaust gases, to lower the running costs of local scrubbers, and to reduce the air pollutant emissions. The EBARA dry scrubber test, performed on a TEL-8S furnace, resulted in DRE values >99% for exhaust gases TEOS, C2H4OH, CIF4, HG, HCI, SiH4, and AsH3. A DRE value >99% was assigned when the outlet gas concentrtion is below the detection limit of the FTIR used (optical path length = 10m).
This abatement efficiency project resulted in a cost reduction estimated by Winbond to be >US$1M annually because of the increased cartridge lifetime for the dry scrubbers. Insurance premiums were also reduced as a result of the lower risk associate with process gases. The exact amount attributed to these efforts has not been determined, however, because the company-wide insurance premium depends on the overall ESH performance of all fabs.
References
1. A.D. Johnson, W.R. Entley, P.J. Maroulis, "Reducing PFC Gas Emissions from CVD Chamber Cleaning," Solid State Technology, p. 103, Dec. 2000.
2. M. Hayes, K. Woods, "Treating Semiconductor Emissions with Point-of-use Abatement Systems," Solid State Technology, p. 141, Oct. 1996.
3. US Environmental Protection Agency (EPA) Test Method 320. "Measurement of Vapor Phase Organic and Inorganic Emissions by Extractive Fourier Transform Infrared (FTIR) Spectroscopy."
4. S-N Li, J-N Hsu, G-H Leu, K. Tang, J. Chiu, "Improved Technique for Evaluating Point-of-Use Abatement Systems," Semiconductor Fabtech 14th Edition, June 2001.
5. D.R. Burgess Jr., M.R. Zachariah, W. Tsang, P.R. Westmoreland, "Thermochemical and Chemical Kinetic Data for Fluorinated Hydrocarbons," Prog. Energy Combust. Sci., Vol. 21, p. 453, 1996.
Shou-Nan Li, Jung-Nan Hsu, and Hui-Ya Shih work at the Center for Environmental, Safety and Health Technology, Industrial Technology Research Institute (ITRI), 11F, Bldg. 51, 195-10 Sec. 4, Chung-Hsing Rd., Chutung, Hsinchu, Taiwan 310, R.O.C. email: [email protected].
Shu-Jen Lin and Jeng-Liang Hong work at Winbond Electronics Corp., No. 9, Li-Hsin Rd., Science-Based Industrial Park, Hsinchu Taiwan, R.O.C.