TECHNICAL SUPPORT FOR THE STACK TESTING OPTION FOR ESTIMATING FLUORINATED GREENHOUSE F-GHG EMISSIONS FROM ELECTRONICS MANUFACTURING FACILITIES UNDER SUBPART I       Office of Air and Radiation, U.S. Environmental Protection Agency
                                  August 2012
                                       
                                       





1. Introduction
This document presents technical information supporting the proposed stack testing option for estimating fluorinated greenhouse gas (F-GHG) emissions from electronics manufacturing facilities required to be reported under subpart I, Electronics Manufacturing, of the Greenhouse Gas Reporting Program (GHGRP). The EPA is proposing the stack testing option as an alternative to using default emission factor (EF)-based methods for estimating F-GHG emissions (see Technical Support for Modifications to the Fluorinated Greenhouse Gas Emission Estimation Method Option for Semiconductor Facilities under Subpart I). The EPA proposed stack testing method would not apply to estimating N2O emissions (see Technical Support for Other Technical Issues Addressed in Revisions to Subpart I for more information on estimating N2O emissions).
This document was developed using information from materials submitted to the EPA by the Semiconductor Industry Association (SIA). Specific stack testing reports with results from tests performed by Intel Corporation, IBM, and Texas Instruments, which are referenced throughout this document, are available in the docket EPA - HQ - OAR - 2011-0028.
2. General Description of and Support for Inclusion of the Stack Test Option
Under other EPA programs and some other subparts under the GHGRP, facilities have the option to estimate emissions using EFs developed by measuring emissions from their stacks over a certain period of time and dividing them by an activity metric (e.g., production) measured over the same period.  This approach, hereinafter referred to as "stack testing" was not previously considered to be a viable option for measuring and estimating greenhouse gas emissions from the electronics manufacturing sector, mainly because of technical issues found with quantifying low concentrations of F-GHGs in stack exhaust streams (Intel, 2010). However, new testing work conducted by SIA members using enhanced Fourier transform infrared spectroscopy (FTIR) techniques showed that gas and fab-specific EFs based on stack testing could be developed and used to estimate F-GHG emissions. Therefore the EPA is proposing to include periodic stack testing as an option for electronics manufacturing fabs to estimate and report F-GHG emissions under subpart I. This method is expected to provide a range of benefits, such as: 
   i)    allowing for estimation of emissions based on periodic, direct measurements of stack emissions from fabs; 
   ii)    reducing industry concerns about exposure of confidential business information (intellectual property) that would be reported to the EPA under other methods; and 
   iii)    possibly reducing compliance costs compared to the use of default EF-based  methods for certain fabs having a limited number of stacks.
The EPA is not proposing to address N2O emissions through a stack testing method, and is instead proposing that fabs continue to rely on default N2O EFs provided in subpart I. This is because a review of SIA stack test data provided to the EPA revealed inconsistent results for stack measurements of N2O emissions for which the cause could not be determined (for N2O stack test data, refer to SIA, 2012g). Due to the inconsistent N2O test results, the EPA concluded that there was not sufficient data to show that stack testing is an appropriate and viable option for estimating N2O emissions.
The remainder of this document discusses the elements of the stack test method, including:
   * Determining which stack systems to test;
   * Determining which F-GHGs to test for;
   * Developing EFs based on stack testing;
   * Estimating Fab F-GHG emissions;
   * Stack testing frequency; and
   * Changes at a fab that would trigger re-testing of stack systems.

3. Determining the Stack Systems to Be Tested
Many fabs have multiple stack systems  that emit F-GHGs. In order to provide accurate fab emission estimates while avoiding excessive burden from measuring emissions from multiple stack systems, the EPA is proposing that a fab would not be required to test certain "low emitting" stack systems for the purpose of developing gas and fab-specific EFs. (Emissions from low-emitting stack systems would be estimated using default EFs provided in subpart I, see Section 6 of this document.) The EPA is proposing to define a low-emitting stack system that is not tested as one that meets all of the following three criteria:
   1. The sum of the F-GHG emissions from all combined stack systems in the fab that are not tested is less than 10,000 metric ton CO2e (mtCO2e) per year; 
   2. Each of the stack systems that are not tested are within the fab's lowest F-GHG emitting stack systems that together emit 15 percent or less of total fab F-GHG CO2e emissions; and 
   3. The F-GHG emissions from each of the stack systems that are not tested can be attributed to only one particular collection of process tools during the test  (i.e., the stack cannot be used as a bypass from other tools that are normally vented through a stack system that does not meet these criteria). 
Each of the elements of this definition and the supporting rationale are described in the following paragraphs.
The EPA is proposing a 10,000 mtCO2e per year absolute threshold for the combined emissions from the low-emitting stack systems that are not tested at a fab. This threshold was selected so that testing would be focused on the largest emitting stack systems. The EPA is proposing that the 10,000 mtCO2e threshold not be applied to individual stack systems because some fabs have large numbers of stack systems with relatively low F-GHG emissions from each system. Therefore, setting a threshold on total fab emissions as opposed to individual stack system emissions ensures that a relatively large portion of the F-GHG emissions are estimated using EFs based on direct fab measurements.
The EPA considered similar issues in establishing a threshold (10,000 mtCO2e) in subpart L, Production of Fluorinated Gases, for lower-emitting process vents.  For such vents, subpart L allows facilities to calculate emissions based on EFs developed through a facility-specific engineering assessment instead of through direct measurements. Please refer to the subpart L Technical Support Document (Technical Support Document for Emissions from Production of Fluorinated Gases: Final Rule for Mandatory Reporting of Greenhouse Gases, November 2010) available in docket EPA-HQ-OAR-2009-0927, for more information on the similar issues considered in selecting the threshold adopted in subpart L.
Because emissions from different fabs may vary widely depending on the processes used, the types of products manufactured, and the extent to which F-GHG abatement systems are installed, the EPA is proposing to also limit the total emissions from the low-emitting stack systems that are not tested to no more than 15 percent of total CO2e F-GHG emissions from the fab.  This relative threshold ensures that a fab's highest emitting stack systems are being tested while reducing the burden of testing stacks that account for a small share of total facility emissions.  In addition, because F-GHG emissions from low-emitting stack systems may be more dilute than those from higher-emitting systems, the provision avoids technical concerns that may be associated with detecting and measuring these emissions. 
The third criterion presented above, that the stack system cannot be used as a bypass for emission from other tools at the fab, is an essential part of defining low-emitting stack systems because it ensures that emissions that are measured during a stack test are representative of a consistent configuration of the fab's exhaust system. Under this definition, emissions from stack systems that are tested could not be diverted, during testing, to stacks that are exempt from testing.  This provision ensures that together, the stack test and the default EF approach for low-emitting stack systems capture emissions from all of the F-GHG emitting tools in the fab.  
Once testing was completed, exhaust systems could be reconfigured without reducing the accuracy of emission estimates as long as one of the following conditions was met : (1) tools and their emissions were tracked according to the configuration of the exhaust system at the time of the stack test (that is, the gas consumption of the tools that were associated with the tested stacks would be multiplied by the measured stack EFs even if those tools were later exhausted to the low-emitting stack system or vice versa), or (2) reconfigurations in exhaust systems were limited to a small percentage of the tools served by each system.  Either of these conditions would ensure that the stack EFs remained representative of the emissions from the tools to which they were applied.  
In order to estimate emissions for purposes of determining whether a stack system is low-emitting, the EPA is proposing that fabs use a method based on the 2006 Intergovernmental Panel on Climate Change (IPCC) Tier 2a method for the electronics manufacturing industry. A Tier 2a method in electronics manufacturing relies on default gas-specific EFs (i.e., kg gas emitted per kg of gas used) and estimates of gas consumption. The method also allows the option to take into account the use of abatement systems. This approach minimizes reporting burden to industry because it does not require allocation of gas consumption between process types or sub-types (e.g., etch and chamber clean), as is required for other default-EF based methods in subpart I. 
In using the Tier 2a method the  EPA is proposing that electronics manufacturers rely on 1) Tier 2a EFs  specific to each type of electronics product manufactured, (see Section 6 for more details), 2) historical estimates of abatements system destruction or removal efficiency (DRE), fab uptime, and gas consumption, 3) gas apportioning factors for stack systems that are based on the number of tools associated with each stack system (i.e., the fraction of an F-GHG apportioned to a stack system is equal to the fraction of all tools using that F-GHG that are vented to that stack system), and 4) global warming potentials of gases usedas inputs to the equations to calculate emissions. 
See Section 6 of this memo for a discussion of how the EPA is proposing that facilities would estimate emissions from low-emitting stack systems for reporting purposes. 
4. Determining Which F-GHGs to Test
   
   4.1. Input Gases
In developing the proposed stack testing method, there was a need for a mechanism to address F-GHGs that are used infrequently. This is because it may be difficult (and perhaps impossible) to schedule stack testing for a time when the fab is using every gas that it uses at some point during the reporting year. The EPA is therefore proposing that fabs have the option to exclude from stack measurements any F-GHG that meets the definition of an intermittent low-use F-GHG as presented below.  
The EPA is proposing to define an intermittent low-use F-GHG as one that meets all of the following three criteria:
   1. The F-GHG is used by the fab but was not used during the period of actual stack testing;
   2. The emissions of that F-GHG do not constitute more than 5 percent of the total annual F-GHG emissions from the fab on a CO2e basis; and
   3. The sum of all F-GHGs that are considered intermittent low-use F-GHGs does not exceed 10,000 metric tons CO2e for that year.
During stack tests, fabs would not be required to measure the F-GHGs that meet all three of these criteria. 
Within criterion 3 above, the EPA is proposing a maximum allowance of 10,000 mtCO2 per year for all intermittent low-use F-GHG combined. This is consistent with the proposed maximum allowance for low-emitting stacks (see Section 3, above). 
The EPA is proposing that fabs use the Tier 2a method as described Section 3 of this memorandum to estimate emissions for the purpose of determining whether an F-GHG is considered to be an intermittent low-use F-GHG by meeting criterion number 2, above.
As part of defining intermittent low-use F-GHGs, the EPA set the relative threshold in criterion number 2 to ensure that the majority of F-GHG emissions are estimated based on actual fab measurements under the stack test method. The EPA evaluated various options for setting the relative threshold, including 15 percent, 10 percent, 5 percent, and 3 percent in combination with the 10,000 mtCO2e absolute threshold (criterion number 3, above). To select the proposed threshold, the EPA estimated fab emissions using gas consumption data submitted by SIA (SIA, 2012a), the IPCC Tier 2a EFs (IPCC, 2006), and assuming no abatement. The results of this analysis are presented in Table 1 below. 
Table 1. Analysis of Intermittent -Low Use F-GHG Relative Thresholds (Source: SIA, 2012a; and IPCC, 2006)


                                 15% Threshold
                                 10% Threshold
                                 5% Threshold
                                 3% Threshold
                             Fab, Year, Wafer Size
                             Number of Gases Used
                                  Number of 
                          F-GHG Required to Be Tested
                   Percent of Emissions from Low Use F-GHGs
                                  Number of 
                          F-GHG Required to Be Tested
                   Percent of Emissions from Low Use F-GHGs
                                  Number of 
                          F-GHG Required to Be Tested
                   Percent of Emissions from Low Use F-GHGs
                                  Number of 
                          F-GHG Required to Be Tested
                   Percent of Emissions from Low Use F-GHGs
Fab A, 2009, 300 mm
                                       7
                                       2
                                      9%
                                       2
                                      9%
                                       2
                                      9%
                                       4
                                      1%
Fab A, 2010, 300 mm
                                       8
                                       2
                                      6%
                                       2
                                      6%
                                       2
                                      6%
                                       2
                                      6%
Fab B, 2009, 200 mm
                                       9
                                       2
                                      13%
                                       2
                                      13%
                                       3
                                      5%
                                       3
                                      5%
Fab B, 2010, 200 mm
                                       9
                                       2
                                      17%
                                       2
                                      17%
                                       3
                                      8%
                                       4
                                      5%
Fab C, 2009, 200 mm
                                       8
                                       2
                                      17%
                                       2
                                      17%
                                       4
                                      2%
                                       4
                                      2%
Fab C, 2010, 200 mm
                                       8
                                       3
                                      21%
                                       4
                                      9%
                                       5
                                      3%
                                       5
                                      3%
Fab E, 2009, 300 mm
                                      10
                                       2
                                      12%
                                       2
                                      12%
                                       3
                                      6%
                                       3
                                      6%
Fab E, 2010, 300 mm
                                       9
                                       2
                                      12%
                                       2
                                      12%
                                       3
                                      5%
                                       4
                                      2%
Note: The analysis in Table 1 is theoretical. The percentage of emissions comprised of intermittent low-use F-GHGs is uncertain because fabs would be required to measure for all F-GHG that are being used as input gases during the test, and whether or not a particular input gas is used during the test depends on actual fab manufacturing.  Therefore, the percentages in Table 1 represent the maximum percentage that could be comprised of F-GHG meeting the definition of intermittent low-use F-GHGs.
In reviewing the results, it is apparent that a 15 percent threshold could result in testing of only two F-GHGs in all but one of the analyzed facilities; in one case, the facility would test for three F-GHGs. Each facility analyzed typically used a total of seven to 10 F-GHGs, as shown in Table 1. The 15 percent threshold would also result in a significant portion of the potential emissions (six percent to 21 percent, based on SIA-provided F-GHG consumption distributions) being represented by F-GHG potentially meeting the definition of intermittent low-use F-GHG and not subject to mandatory testing during the fab-specific stack measurements.  Reducing the threshold from 15 percent to 10 percent would not significantly increase the number of F-GHGs that would be tested or reduce the percentage of potential emissions from intermittent low-use F-GHGs. Changing to a five percent threshold for individual F-GHGs would reduce the potential percentage of total emissions from intermittent low-use F-GHGs to less than 10 percent for all fabs covered in Table 1, while only moderately increasing the burden associated with the minimum number of F-GHGs to be measured. Decreasing the threshold from 10 percent to five percent would increase the minimum number of F-GHGs to be tested from two F-GHG to three F-GHG at five of the facilities analyzed. Using a three percent threshold for individual F-GHGs would drop the percentage of potential emissions from intermittent low use F-GHGs to no more than six percent while increasing the minimum number of F-GHGs to be tested from two to four F-GHGs at one facility, and from three to four F-GHGs at two facilities. However, under a three percent threshold, a fab would be required to measure gases that may be used only rarely, presenting significant scheduling hurdles.
The proposed relative threshold of less than five percent of the total estimated emissions for the definition of intermittent low-use F-GHGs would result in actual testing of greater than 90 percent of overall emissions from tested stack systems. In terms of the minimum number of F-GHGs tested, this would represent (on average) about one-third of the total number of F-GHGs used, but would increase technical feasibility compared to testing all F-GHGs. As noted above, the actual number of F-GHGs tested would depend on which F-GHGs are actually being used when the stack testing is conducted. 
If a F-GHG does not meet the proposed definition of an intermittent low-use F-GHG, and was not used during the emissions test, an additional stack test would need to be conducted in the reporting year during a period when that gas is being used. However, if a F-GHG that was not an intermittent low-use F-GHG is no longer used, re-testing would not be required, and F-GHG emissions would be calculated according to the process for intermittent low use F-GHGs.
See Section 6 for a discussion on how the EPA is proposing to estimate emissions of the intermittent low-use F-GHGs for reporting purposes.
   3.1. By-Products
In addition to identifying input gases to test for in stack systems it is important to identify known by-products of processes used at the fab that will also have to be tested for. By-product emissions are estimated to make up from 14 to 19 percent of emissions at 200 mm fabs and from 14 to 24 percent of emissions at 300 mm fabs.  Table 2 below shows a survey of known by-products from electronics manufacturing.
          Table 2. Common Electronics Manufacturing By-Product Gases
                                    Source
                            Identified By-Products
2006 IPCC Guidelines for National Greenhouse Gas, Volume 3, Chapter 6 
CF4, C2F6, C3F8, CHF3
International SEMATECH Manufacturing Initiative Environmental Safety and Health Technology Center Etch Process Equipment Emissions Characterization Data (SIA, 2012b)
CF4, C2F6, C4F8, C4F6, CHF3
Response to EPA's Stack Test Question 1 (SIA, 2012g)
CF4, C2F6, C3F8
In reviewing Table 2, it can be seen that CF4 and C2F6 are common by-products identified across all sources surveyed. Therefore, the EPA is proposing that all fabs using the stack testing method specifically test for CF4 and C2F6, even in situations when they are not used as input gases (i.e., are emitted only as by-products).  In addition to proposing that these testing always include CF4 and C2F6, the EPA is also proposing that fabs analyze and document any expected by-products (other than CF4 and C2F6) for the upcoming reporting year. The EPA is proposing that any identified by-products be tested for as well during the reporting year. Expected by-products would be identified  by an  analysis of any possible F-GHG  by-products formed from the input gases (F-GHGs) used in the previous reporting year and expected to be consumed in the current reporting year. When developing this analysis the EPA is also proposing that fabs consider at a minimum the by-products identified in Tables I-3 through I-7 of subpart I (or those identified in the second row of Table 2, found in SIA, 2012b).
Another option that the EPA is considering is to require testing for all F-GHGs that have been identified as by-products of any input gas in previous testing throughout the electronics industry.  This set would include C3F8, C4F6, C4F8, and CHF3 in addition to CF4 and C2F6. The EPA is considering this option because the identities and quantities of by-products generated at a particular facility at a particular time can be difficult to predict, and the costs of testing for additional by-products are expected to be modest.  In the one set of semiconductor facility stack tests that tested for the full range of potential by-products listed above, a F-GHG by-product was found for which there are no by-product formation factors listed in Tables I-3 and I-4 (the default EFs tables applicable to semiconductor facilities).  This was C3F8, which accounted for 40% of the GWP-weighted by-product emissions of the fab (2% of the total GWP-weighted emissions) in one of the two tests at the fab.  (In the other test, the C3F8 emissions accounted for a much smaller, but still measurable, share of emissions.)  If unexpected by-products occur in similar proportions at other facilities, failing to measure for them could lead to routine underestimates of emissions at those facilities.  
Once a method has been validated for all potential by-product gases, the costs of testing for six, as opposed to two, of these gases are expected to be low.  All gases would be tested for at the same time by the same equipment and personnel.  The only additional costs would be those associated with the analysis of the resulting data, which is performed automatically.  It appears very unlikely that these additional costs would exceed 10% of the overall costs of stack testing, and they would probably be considerably less.  
5. Developing Gas-Specific and Fab-Specific Emission Factors 
The proposed subpart I stack testing option would require fabs to develop EFs based on a correlation between measured gas consumption and stack emissions for all stack systems tested and for all F-GHGs tested. The EFs for input F-GHGs would be in the units of kilograms of the F-GHG emitted per kilogram of the F-GHG consumed. The EFs for a by-product F-GHG that is not used as an input F-GHG would be in units of kilograms of the by-product F-GHG emitted per kilogram of all F-GHG consumed as input gases combined. If an input F-GHG's emissions exceeded its consumption, then the emissions in excess of consumption would be treated as if they were by-product F-GHG emissions, except that the denominator in the EF would not include the consumption of that F-GHG (See Section 5.3).
   5.1. Measuring Stack F-GHG Emissions
This section addresses the provisions related to stack measurements of F-GHGs, including measurement methods, durations of testing, field detection limits, and treatment of emission concentrations below the field detection limits.
The proposed stack testing option would require fabs to measure F-GHG emissions for all F-GHG used as inputs to processes and for any expected by-product F-GHGs, except for F-GHG emitted from low-emitting stack systems and for F-GHG that meet the definition of intermittent low-use gases. 
      5.1.1. Measurement Methods
The calculation of fab EFs are based on (1) the measurements of the concentrations of F-GHGs in the stack systems and (2) on the measurement of the stack flows. 
The EPA is proposing that EPA Method 320 be used to measure the concentration of F-GHGs in the stack systems using FTIR. Please refer to Method 320 for detailed information about the measurement method and procedures. Please note that in utilizing the work performed by SIA member companies (Intel, IBM, and Texas Instruments) in consultation with the EPA, the use of this method for stack testing to measure F-GHG emissions at electronics manufacturing facilities has been validated using Section 13 of Method 320.  SIA member companies provided EPA with timely stack testing plans and reports at the beginning of the process that were invaluable to evaluating Method 320 at these facilities.  
For the measurement of the stack gas volumetric flow rates, the EPA is proposing to use Methods 1 or 1A, and Methods 2, 2A, 2C, 2D, 2F or 2G. These methods are based on the measurement of the gas velocity and also require the measurement of the gas molecular weight using Methods 3, 3A, or 3B, and measurement of the stack's moisture content using FTIR or Method 4. Please refer to Table I-9 in subpart I for more information and to the links in the footnotes within this paragraph for more details on the proposed methods for determination of the stacks' flow rates.
For supporting data and information on the proposed use of the methods described above in stack testing, please refer to Intel Corporation, IBM, and Texas Instruments test reports available in the EPA docket EPA - HQ - OAR - 2011-0028. Through testing, it has been shown that the identified methods are viable options for measuring F-GHG emissions from semiconductor (and other electronics) manufacturing. 
The EPA is also proposing to allow fabs to use alternative measurement methods for stack system testing to provide flexibility in the rule. This is consistent with the flexibility provisions provided to reporters in other subparts, such as subpart L (see Mandatory Reporting of Greenhouse Gases: Additional Sources of Fluorinated GHGs, published December 2010, (75 FR 74774)). An alternative method is any method of sampling and analyzing for an F-GHG that is not a method specified in subpart I. The EPA is proposing that alternative test methods would have to be validated according to EPA Method 301 (Field Validation of Pollutant Measurement Methods From Various Waste Media). Prior to using an alternate test method, the facility would be required to submit the method along with a proposed test plan to the EPA for review and approval by the EPA. This process would ensure that any method outside of those specified in subpart I is viable and appropriate. Once the EPA approves an alternate method, the method could be used by other fabs that meet the same conditions under which the alternative method was approved. The EPA would specify those conditions in the approval of the alternative method.
      5.1.2. Maximum Field Detection Limits
In using Method 320, maximum field detection limits (FDLs) need to be specified because the FDLs achieved by a method and analytical instruments can have a significant impact on the quality of the measurements. The maximum FDLs are the limits at which an F-GHG should be detectable when the method is conducted properly and the analytical instruments are used correctly and of reasonable quality. If the FDL for an F-GHG were so high that it represented a relatively large fraction of the fab's emissions of that F-GHG, the uncertainty of the resulting emission estimate would be correspondingly high. Therefore, the EPA is proposing maximum FDLs for stack tests for electronics manufacturing facilities. Fabs using the stack testing option would need to achieve FDLs at or below the maximum FDLs in order for the stack testing results to be acceptable for estimating F-GHG emissions. The FDL would be determined for each stack test for each analyte by the person conducting the stack test.  The proposed maximum FDLs, in parts per billion by volume are presented in Table 3 below.
  Table 3. Proposed Maximum Field Detection Limits (Table I-10 in Subpart I)
                                    Analyte
                                      CF4
                                     C2F6
                                     C3F8
                                     C4F6
                                     C5F8
                                     C4F8
                                     CH2F2
                                     CH3F
                                     CHF3
                                      NF3
                                      SF6
                         Other fully fluorinated  GHGs
                            Other fluorinated GHGs
                              Maximum FDLs (ppbv)
                                       5
                                       5
                                       5
                                       5
                                       5
                                       5
                                      10
                                      10
                                       5
                                       5
                                       1
                                       5
                                      10
The proposed maximum FDLs are based on (1) the EPA's review of the FDLs that have been achieved at three different semiconductor facilities (please refer to Intel Corporation, IBM, and Texas Instruments test reports available in docket EPA - HQ - OAR - 2011-0028), and (2) an analysis of the magnitude of the emissions that would occur (in CO2e) at various possible maximum FDLs. (The latter provides an indication of the uncertainty of emission measurements using methods and analytical instruments with those FDLs.) The results of this review and analysis are presented in Table 4, below.
Table 4 presents the EPA's analysis of how the proposed maximum FDLs translate into corresponding F-GHG emissions at the proposed maximum FDLs. To determine these emissions, the EPA used the known molecular weights for each gas, an assumed fab-wide stack flow rate, and simple unit conversions. 

        Table 4. Emissions at Proposed Maximum Field Detection Limits 
                                    Analyte
                                   Maximum 
                                  FDL (ppbv)
                           Molecular Weight (g/mole)
                       Assumed Fab-Wide Flow rate (scfm)
                                    kg/min
                                     kg/yr
                                     mt/yr
                                      GWP
                                   mtCO2e/yr
CF4
                                       5
                                                                             88
                                                                        150,000
                                                                       7.77E-05
                                      41
                                     0.04
                                                                           6500
                                                                           266 
C2F6
                                       5
                                                                            138
                                                                        150,000
                                                                       0.000122
                                      64
                                     0.06
                                                                           9200
                                                                           590 
C3F8
                                       5
                                                                            188
                                                                        150,000
                                                                       0.000166
                                      87
                                     0.09
                                                                           7000
                                                                           611 
C4F8
                                       5
                                                                            200
                                                                        150,000
                                                                       0.000177
                                      93
                                     0.09
                                                                           8700
                                                                           808 
CH2F2
                                      10
                                                                             52
                                                                        150,000
                                                                       9.19E-05
                                      48
                                     0.05
                                                                            675
                                                                            33 
CHF3
                                       5
                                                                             70
                                                                        150,000
                                                                       6.18E-05
                                      33
                                     0.03
                                                                          11700
                                                                           380 
NF3
                                       5
                                                                             71
                                                                        150,000
                                                                       6.27E-05
                                      33
                                     0.03
                                                                          17200
                                                                           567 
SF6
                                       1
                                                                            146
                                                                        150,000
                                                                       2.58E-05
                                      14
                                     0.01
                                                                          23900
                                                                           324 
Total
 
 
 
 
 
 
 
                                                                         3,578 
In reviewing how the proposed FDLs for specific gases translate into emissions it can be seen that the proposed maximum FDLs are lower for F-GHGs that have higher GWPs and are easier to detect (e.g., SF6, for which a maximum FDL of 1 ppbv is proposed) and higher for F-GHGs that have lower GWPs and are more difficult to detect (e.g., CH3F, for which a maximum FDL of 10 ppbv is proposed). For perfluorocarbons and CHF3, the EPA is proposing maximum FDLs of 5 ppbv. The proposed maximum FDLs are generally, though not always, close to the average FDLs achieved across all three facilities that submitted FDL information to the EPA (see IBM, Intel, and Texas Instruments test reports in docket EPA - HQ - OAR - 2011-0028). All of the maximum FDLs are higher than the FDLs achieved at the two facilities that used enhanced FTIR during their stack testing. Emissions of F-GHGs at the maximum FDLs generally range between 200 and 800 mtCO2e per gas, based on a total flow rate at the fab of 150,000 standard cubic feet per minute (scfm), totaling approximately 3,600 mtCO2e for commonly used F-GHGs with GWPs in Table A-1. Thus, the maximum FDLs are readily achievable using proven detection technology but are low enough to ensure quality data are used to develop stack EFs.  (It is important to note that at an actual facility comparable to the example represented by Table 4, the most commonly used F-GHGs would be measured in concentrations that are well above the proposed maximum FDL, and the total emissions of the F-GHG that are at or below the proposed maximum FDL would be substantially lower.) 
As seen in Table 3, the EPA is also proposing maximum FDLs for two broad categories of F-GHGs, "Other fully fluorinated GHGs" and "Other fluorinated GHGs". The EPA is including these categories because not all known F-GHGs used in electronics manufacturing (e.g., C4F8O) were tested as part of the  stack testing used to develop this proposed stack method still other F-GHGs may be used or emitted in the future.  The maximum FDL proposed for "fully fluorinated GHGs" is 5 ppbv, the same as that proposed for all the PFCs that have been tested to date.  Like PFCs, other fully-fluorinated GHGs are relatively potent GHGs and are likely to be detectable in small quantities using FTIR.   The maximum FDL proposed for "other fluorinated GHGs" is 10 ppbv, the same as that proposed for CH2F2 (HFC-32).  Like HFC-32, other fluorinated GHGs are likely to be less potent than fully fluorinated GHGs and somewhat more difficult to detect.   
      5.1.3. Test Durations and Conditions
The EPA is proposing that stack tests be performed for a minimum of 8 continuous hours. For four different fabs the difference between the upper and lower concentrations over an 8 hour difference at the 95-percent confidence interval was typically less than 15 percent (SIA, 2011). Also, the difference in day-to-day flow rates over 4 to 5 days was shown to be typically around one percent. Finally, a continuous measurement of CF4 emissions during a 24 hour period showed that the mean concentration was 167 ppbv, while the lower and upper concentrations were 162 and 172 ppbv respectively (95-percent confidence interval). Over the same 24 hour period, the 8-hour running average ranged between 144 and 186 ppbv, yielding similar results to a 24-hour sample time. Based on these results (from SIA, 2011), the EPA determined that the proposed 8-hour test duration would appropriately capture short term variations in stack emissions.
The EPA is also proposing that facilities adhere to two conditions while conducting stack testing in order to ensure that EFs developed from stack tests are representative of fab emissions. First, while the EPA does not require that emissions from all stack systems be tested simultaneously, the EPA is proposing that there be no changes between tests in the stack flow configuration (i.e., the configuration of the ducts between sets of process tools and any connected point-of-use (POU) abatement systems and their corresponding waste streams that are ultimately vented through the stack) during the entire testing period. The EPA is also proposing that there be no changes in the centralized abatement systems during testing; if any are present (the EPA's current understanding is that centralized abatement systems are rarely used in electronics manufacturing, if at all). This would ensure that emissions from each tool are measured by one and only one test (i.e., no emissions are missed or double counted). Second, the EPA is proposing that stack tests be conducted for a period during which the fab is operating at a representative operating level, and with the POU abatement systems connected to the stack being tested operating with at least 90 percent uptime during the 8-hour (or longer) period or at no less than 90 percent of the average uptime measured during the previous reporting year. This requirement would ensure that the EFs developed from stack tests are representative of fab operations over the reporting year.


      5.1.4. Treatment of F-GHG Concentrations That are Below the FDL
In performing stack measurements, fabs may have cases in which certain input F-GHGs (excluding those that are classified as intermittent low-use gases) or emitted as by-product F-GHGs cannot be detected or are not continuously detected. For these situations, the EPA is proposing that:
   * If an F-GHG is consumed during testing, but emissions are not detected, the reporter would use one-half of the FDL for the concentration of that F-GHG in calculations.
   * If an F-GHG is consumed during testing and detected intermittently during the test run, the reporter would use the detected concentration for the value of that F-GHG when available and use one-half of the FDL for the value when the F-GHG is not detected.
   * If an F-GHG is not consumed during testing but is detected intermittently as a by-product gas, the reporter would use the measured concentration when available and use one-half of the FDL for the value when the F-GHG is not detected.
   * If an F-GHG is an expected by-product gas (e.g., CF4, C2F6) of the stack system tested and is not detected during the test run, the reporter would use one-half of the FDL for the value of that F-GHG.
   * If an F-GHG is not used, is not an expected by-product of the stack system, and is not detected, then the reporter would assume zero emissions for that F-GHG for the tested stack system.
The EPA is proposing to use half the FDL (note: this is the FDL as defined in EPA Method 320 and determined for each stack test, not the proposed maximum FDL), for these cases because the EPA expects that the proposed treatment of these non-detect values will avoid any potential under-counting of any F-GHG that are expected to be in the emissions from a given process and F-GHG input gas combination. At the same time, the proposed treatment will provide a reasonable estimate of emissions of F-GHG that occur in concentrations that are below the FDL. The EPA's analysis of testing data provided by SIA has shown that emission measurements of gases known to be used and for which the concentration was below the FDL accounted for about 0.1 percent of F-GHG consumption and would account for about 0.1 percent of emissions on a CO2e basis if the concentration was assumed to be one-half of the FDL as outlined above. The measurements were at done at Texas Instruments and Intel fabs. At Texas Instruments, emissions of c-C4F8 (an input gas) were below detection; at one-half the FDL, emissions would be 65 mtCO2e per year and would represent 0.1 percent of total estimated facility emissions. At Intel, emissions of CH2F2 (also an input gas) were below detection; at one-half the FDL, emissions would be 132 mtCO2e per year and would also represent 0.1 percent of total estimated facility emissions. 
   5.2. Measuring Fab Activity (F-GHG Consumption)
The EPA is proposing that gas consumption be used as the metric of fab activity to develop fab-specific EFs from stack testing. Other potential metrics of fab activity, such as wafer movements, were considered; however, a strong correlation between gas consumption and emissions has been shown over time (SIA, 2011 and SIA, 2012c).
Data supplied by SIA for multiple fabs have shown that gas consumption is consistently linear over time and that emissions are also a linear function of gas consumption (SIA, 2011 and SIA, 2012c). SIA provided the EPA with F-GHG consumption from four fabs for several different F-GHGs. For CF4 and CHF3, consumption data were available from three fabs, and for SF6 and NF3, data were available from four fabs. For all four of these F-GHGs, consumption data were available for time periods measured in hours, days, and months.  These data showed that consumption of these four F-GHG was linear over all of the time periods measured, including 8-hour stack testing periods, periods of several days to weeks, and, at one fab, annual 12-month calendar periods. Consumption data over a period of several hours were also available for C2F6 from one fab, and consumption data over a period of a few weeks were available for C4F6, C4F8, and CH2F2 from another fab. These data also showed that consumption was linear for these gases at these two fabs over the time periods measured. (SIA, 2012c). 
Based on these results, the EPA concluded that gas consumption was a good measure of fab activity, and that stack EFs could be expressed as a function of F-GHG consumption. Furthermore, the EPA concluded that these EFs based on monitoring emissions and F-GHG consumption for short time periods would be representative of longer term emissions such that they could be applied annually. 
However, there were some cases in the SIA data where the short-term correlation between consumption and emissions for CF4, SF6, and NF3 could not be replicated for other F-GHGs during testing at fabs with limited consumption of the other F-GHG because the scales used in production for such gases could not resolve their weight change across the 8-hour test period (SIA, 2012c). Thus, for gases for which it may be difficult to accurately measure short-term (i.e. 8-hour) consumption, the EPA is proposing that reporters would use one or more of the following measures to increase the accuracy of their consumption estimates, if needed:
   1) Draw from single gas containers (instead of from multiple containers attached to a common manifold). 
   2) Calculate consumption from prorated long-term (e.g., monthly) consumption data.
   3) Increase the length of the test period above 8 hours.
 The EPA is proposing various options for measuring gas consumption during the test period:
   1) Fabs could use measurement equipment (e.g., gas flow meters, weigh scales, or pressure gauges) to monitor F-GHG consumption. 
   2) Gas consumption calculated based on pressure measurements would be corrected for temperature and non-ideal gas behavior according to the following:
         a) If a reporter measured gas consumption using pressure the reporter would either measure the temperature of the F-GHG container when the sampling periods begin, end, and when containers are changed out, or measure the temperature every hour. 
         b) Reporters would convert the measured pressures to masses using an appropriate equation of state and the measured temperatures.
It is critical that gas consumption measurements based on changes in cylinder pressure be corrected for non-ideal gas behavior because measurements of gas consumption may sometimes be affected by temperature fluctuations that change gas pressure, especially with gas cylinders with low drawdown rates. These issues may have a significant impact on the short-term gas consumption measurements used to establish EFs (SIA, 2012e).
   2.1. Estimating Abatement System Downtime
For fabs that have F-GHG abatement systems associated with stack systems that are tested, it is critical to consider appropriately the downtime of those abatement systems for two purposes: 1) when developing stack EFs, and 2) when estimating emissions using stack-testing based EFs. For example, if the uptime of abatement systems is 100 percent during the period of time in which stack emissions are measured for developing an EF, yet average uptime for the fab for the reporting year is 85 percent, using an EF based on conditions where uptime is 100 percent would underestimate F-GHG emissions because it would not account for excess emissions during the time that the abatement system is not operating. Therefore, the EPA is proposing methods for estimating the downtime of abatement systems at a fab that uses stack testing and accounting for the excess emissions. 
To determine downtime for the purpose of developing stack EFs, the EPA is proposing that the amount of abatement system downtime be estimated using the number of tools, the total duration of the stack testing period, and the duration of the abatement system downtime during the testing. For example, if five POU abatement systems are down for times of 10, 15, 25, 30, and 40 minutes during an 8-hour test, the total POU system downtime would be 120 minutes, or 5.0 percent of the total possible abatement system and tool operating time for the five tools (2,400 minutes).  Using these data and the average DRE for the POU abatement systems, the EF measured during the testing would be adjusted to an EF representing POU abatement systems with 100 percent uptime (zero percent downtime).
The EPA is proposing that the downtime measured over the year would be used to determine an uptime factor that would be an aggregate for all abatement systems in the fab, and that would be calculated using proposed Equation I-23 in subpart I. This value would be used for calculating annual fab emissions. Abatement system downtime would be considered any time during which the abatement system was not installed, operated or maintained according to the manufacturer's specifications. The fab's operator would determine the sum of the downtime for all abatement systems during the year, and divide this sum by the sum of the possible annual operating time for each of the tools connected to those abatement systems in the fab to determine the downtime fraction. The downtime fraction would be the decimal fraction of operating time that the abatement systems were not installed, operated, and maintained according to the manufacturer's specifications. The uptime fraction used in the emissions calculations would be equal to 1 minus the downtime fraction. 
The total possible annual tool operating time would be calculated by assuming that tools that were installed for the whole of the year were operated for the entire year. The total possible tool operating time would be prorated to account for the days in which a tool was not installed; any partial day that a tool was installed would be treated as a full day of tool operation. For an abatement system with more than one connected tool, the tool operating time would be equivalent to a full year if at least one tool was installed at all times throughout the year. The fab would also be able to account for time that tools are idle and no gas is flowing through the tools to the abatement system.
The method outlined above is different than the method for estimating uptime of abatement systems for the default EF based method, which requires an uptime to be determined for each gas and process type combination as opposed to calculating downtime and uptime based on an aggregate for all tools in the fab. This is because the default EF method requires emissions to be estimated based on process types (etch and CVD chamber cleaning), whereas the stack testing method looks at emissions by stack systems as opposed to by process type. Allowing the proposed simplified approach for calculating downtime for stack testing would reduce the burden on fabs that elect to do stack testing because it does not require tracking downtime for each gas and process type.
To ensure that the simplifying assumption (i.e., the operation of abatement systems is consistent across gases, tools, and processes) would not drastically impact emission estimates, the EPA performed an analysis to compare fab emissions if downtime is assigned to each type of tool (etch or chemical vapor deposition) and fab emissions if downtime is assumed to be evenly distributed among all tools. The difference in emissions as calculated both ways in this analysis varied depending on the fraction of tools abated and the downtime assumed, increasing with both.  Where the fraction of tools abated was assumed to be 100 percent and the average downtime was assumed to be 35 percent (both likely to be maximum values), the difference was about 5 percent of total (abated) emissions and about 3 percent of total unabated emissions.   (The error as a fraction of unabated emissions was calculated to permit objective comparisons between the errors at highly abated facilities with low absolute emissions and less abated facilities with high absolute emissions.)  Further, the simplifying assumption is not expected to have a substantive impact on overall facility emission estimates because abatement system uptime is generally high as seen by a sampling of fabs that submitted uptime data to the EPA; for example around or above 90 percent (SIA, 2012f). 
   2.2. Calculating Emission Factors
The EPA is proposing that an EF be calculated for each F-GHG based on the aggregated results from all the tested stack systems at a fab. (These EFs would not be calculated for F-GHGs that are intermittent low-use F-GHGs.) The EFs would be based on stack emissions measurements (kg F-GHG emitted) and gas consumption (kg of F-GHG consumed) for all higher emitting stack systems during the period of testing (e.g., 8 hours) using the calculation methods described in the proposed regulatory text. The result is an EF for input gases in the units of kg of F-GHG emitted per kg of F-GHG consumed, taking into account abatement system use and performance. For by-product F-GHG emissions that are not also used as input F-GHG, the EPA is proposing that the fab-specific by-product EF be calculated as the mass of the by-product F-GHG emitted divided by the summed masses of all the F-GHGs consumed; the calculation methods for which are also presented in the proposed regulatory text. 
The EF that is determined from the stack testing results would be corrected to account for any abatement system downtime that occurs during the actual stack testing. As described in Section 5.3 of this memo, the fab would correct the measured emissions to account for the abatement system downtime, and would use either default or measured DREs for the abatement system to calculate gas and fab-specific EFs that reflect 100 percent uptime of all relevant abatement systems. Annual unabated emissions used in accounting for abatement system downtime would be calculated as the measured EF divided by (1  -  DRE).

      2.2.1. Treatment of F-GHGs For Which Emissions Exceed Consumption
In some cases, emissions of a particular F-GHG input gas may exceed consumption of that gas because the F-GHG is generated as a by-product of the other input gases.  This is often the case for CF4.  In these cases, the EPA is proposing that the reporter equate emissions of the input gas to the consumption of that gas for purposes of calculating the F-GHG input EF  using Equation I-19.  This would result in a calculated input EF of 1.0.  The reporter would treat the remainder of the input gas's emissions as a by-product of the other input gases using Equation I-20.  In this instance, the denominator in Equation I-20 would include the consumption of all the F-GHGs other than the F-GHG with the "extra" emissions. This treatment of the denominator reflects the fact that the EPA is assuming that the F-GHG in the numerator is formed as a by-product from all other F-GHGs, while the emissions from the actual consumption of that F-GHG as an input are being accounted by proposed Equation I-19.
For example, if the fab consumed 100 kg of an F-GHG during the testing, but the stack testing measured 300 kg of that gas, the reporter would treat 100 kg of that F-GHG's emissions as an input gas in proposed Equation I-19, and 200 kg of that F-GHG's emissions as a by-product gas in proposed Equation I-20. 
For calculating emissions from an F-GHG with "extra" emissions, the input portion of emissions would be assumed to equal consumption of that F-GHG, and the by-product portion of emissions would be determined by multiplying the by-product EF by the sum of the consumption of all F-GHGs other than the by-product F-GHG. 
The advantage of this approach is that it reflects the physical mechanism through which emissions of an input gas exceed consumption of that gas.  Because mass is conserved, the emissions of an input gas that are in excess of consumption of that gas must be attributable to the other input gases.  These "extra" emissions are expected to vary with the facility's consumption of the other input gases rather than with the facility's consumption of the emitted gas with the "extra" emissions.  Reflecting this in the by-product EF  would lead to more accurate emission estimates and would help to prevent large swings in EFs that could result when consumption of the gas with the "extra" emissions varies from test to test.  For example, this could help a facility to avoid a 20 percent or greater relative standard deviation in its CF4 EF, which would otherwise prevent the facility from qualifying to skip testing for five years (discussed further below).  
Note that the proposed approach includes a simplification that would in some cases affect the "extra" emissions that are reassigned as by-products of other input gases.  Specifically, Equation I-19 accounts for the application of abatement devices in developing the EF  for each input gas.  This means that if abatement is used at a facility, an EF of 1.0 for a gas with extra emissions at that facility actually equates to a pre-abated EF of greater than 1.0 for that gas.  In other words, some "extra" emissions continue to be included in the input EF for that gas.  
The impact of this simplification on the by-product emissions calculated under Equation I-20 depends on the extent to which a particular input gas is also generated as a byproduct and especially on the overall level of abatement for the stack system tested, i.e., the fraction of tools with abatement systems and the DREs of those systems.  Table 5 shows the differences in the fractions of emissions counted as by-products with and without the simplification, based on a range of by-product generation rates and abatement levels.  (For simplicity, the consumption during the test of the F-GHG with the extra emissions is equated to 1 kg, although this is lower than consumption of most gases is likely to be.)
  Table 5.  Emissions Counted as By-Products with and without Simplification
                           Consumption of F-GHG (kg)
              Emissions of F-GHG over and above consumption (kg)
                         Overall abatement level (a*d)
                               Abated Emissions
                     Fraction "Extra" with Simplification
                              Unabated Emissions
                    Fraction "Extra" without Simplification
       Difference in Fractions as Share of Extra without Simplification
                                                                              1
                                                                            0.5
                                                                             0%
                                                                            1.5
                                                                            33%
                                                                            1.5
                                                                            33%
                                                                             0%
                                                                              1
                                                                            0.5
                                                                            10%
                                                                           1.35
                                                                            26%
                                                                            1.5
                                                                            33%
                                                                            22%
                                                                              1
                                                                            0.5
                                                                            50%
                                                                           0.75
                                                                             0%
                                                                            1.5
                                                                            33%
                                                                           100%
                                                                              1
                                                                            0.5
                                                                            90%
                                                                           0.15
                                                                             0%
                                                                            1.5
                                                                            33%
                                                                           100%
                                                                              1
                                                                            0.5
                                                                            99%
                                                                          0.015
                                                                             0%
                                                                            1.5
                                                                            33%
                                                                           100%
                                                                              1
                                                                              1
                                                                             0%
                                                                              2
                                                                            50%
                                                                              2
                                                                            50%
                                                                             0%
                                                                              1
                                                                              1
                                                                            10%
                                                                            1.8
                                                                            44%
                                                                              2
                                                                            50%
                                                                            11%
                                                                              1
                                                                              1
                                                                            50%
                                                                              1
                                                                             0%
                                                                              2
                                                                            50%
                                                                           100%
                                                                              1
                                                                              1
                                                                            90%
                                                                            0.2
                                                                             0%
                                                                              2
                                                                            50%
                                                                           100%
                                                                              1
                                                                              1
                                                                            99%
                                                                           0.02
                                                                             0%
                                                                              2
                                                                            50%
                                                                           100%
                                                                              1
                                                                              4
                                                                             0%
                                                                              5
                                                                            80%
                                                                              5
                                                                            80%
                                                                             0%
                                                                              1
                                                                              4
                                                                            10%
                                                                            4.5
                                                                            78%
                                                                              5
                                                                            80%
                                                                             3%
                                                                              1
                                                                              4
                                                                            50%
                                                                            2.5
                                                                            60%
                                                                              5
                                                                            80%
                                                                            25%
                                                                              1
                                                                              4
                                                                            90%
                                                                            0.5
                                                                             0%
                                                                              5
                                                                            80%
                                                                           100%
                                                                              1
                                                                              4
                                                                            99%
                                                                           0.05
                                                                             0%
                                                                              5
                                                                            80%
                                                                           100%
                                                                              1
                                                                              9
                                                                             0%
                                                                             10
                                                                            90%
                                                                             10
                                                                            90%
                                                                             0%
                                                                              1
                                                                              9
                                                                            10%
                                                                              9
                                                                            89%
                                                                             10
                                                                            90%
                                                                             1%
                                                                              1
                                                                              9
                                                                            50%
                                                                              5
                                                                            80%
                                                                             10
                                                                            90%
                                                                            11%
                                                                              1
                                                                              9
                                                                            90%
                                                                              1
                                                                             0%
                                                                             10
                                                                            90%
                                                                           100%
                                                                              1
                                                                              9
                                                                            99%
                                                                            0.1
                                                                             0%
                                                                             10
                                                                            90%
                                                                           100%
As shown by Table 5, where stack systems have low abatement levels, the simplification has little impact.  However, where stack systems have very high abatement levels (e.g., 90% or more), the simplification would result in no "extra" emissions being counted as a by-product even when a substantial share (e.g., 90%) of the unabated emissions of an F-GHG actually are formed as a by-product.  
To ensure that all "extra" emissions were reassigned as by-products of other gases, it would be necessary to 
   (1)  Calculate the unabated emissions of the gas with the extra emissions;
   (2) Assign the share of these unabated emissions that equals consumption to Equation I-19; 
   (3) Reapply the abatement level to these emissions in Equation I-19 (reducing the EF from 1.0 to a lower level); 
   (4) Reapply the abatement level to the "extra" emissions that will be treated as a by-product in Equation I-20; and
   (5) Use Equation I-20 to calculate the resulting by-product factor. 
      5.5.1. Emission Factor Variability and Uncertainty
In establishing a stack test method, the EPA took into consideration the sources of potential variability and uncertainty in EFs established at a fab. These included:
   * Variability of F-GHG utilization and by-product generation factors across processes:  
         o Factors can vary over wide ranges, e.g., an order of magnitude or more.  
         o Processes vary as the mix of particular products and their recipes varies. This mix varies over time to different extents at different facilities.
   * Variability in abatement levels: 
         o Either long-term, due to installation or removal of abatement systems, or short-term, due to changes in abatement system uptime.
   * Variability in flow patterns between tools and particular stacks.
   * Uncertainty in gas consumption measurements.
The EPA reviewed EF data developed from stack testing for March and August of 2011 for a fab owned and operated by one company (Company D) that produces a range of semiconductor products and uses an array of F-GHGs. The EFs are presented in Table 6 below. Table 6 shows that the developed stack EFs are generally similar for the fab between the months by gas and for the stack, showing consistency in EFs at different times of the year.

Table 6. Stack Test Emission Factors for a Company D fab in March 2011 and August 2011 (SIA, 2012g)
EF
                                     Units
                                F-GHGs Consumed
                                  Byproducts
                                    Stack 
                                    Mar-11
 
 
                                      NF3
                                      CF4
                                     CHF3
                                      SF6
                                     CH2F2
                                     C4F8
                                     C5F8
                                     C2F6
                                     C3F8
                                       
EF
(kg emit / kg use)
                                    0.0673
                                    0.8669
                                    0.0591
                                    0.3126
                                    0.1108
                                    0.0572
                                    0.0247
                                       
                                       
                                       
bpEF
(kg emit / kg total gas consumed)
                                       
                                       
                                       
                                       
                                       
                                       
                                       
                                    0.0069
                                    0.0004
                                       
% of Total Emissions 
percent
                                     58.01
                                     26.59
                                     8.64
                                     2.08
                                     0.00
                                     0.10
                                     0.00
                                     4.36
                                     0.22
                                       
Stack EF
mtCO2e Emitted/mtCO2e Consumed
                                       
                                       
                                       
                                       
                                       
                                       
                                       
                                       
                                       
                                    0.0954
                                    Aug-11
 

                                      NF3
                                      CF4
                                     CHF3
                                      SF6
                                     CH2F2
                                     C4F8
                                     C5F8
                                     C2F6
                                     C3F8
                                       
EF
(kg emit / kg use)
                                    0.0576
                                    0.8321
                                    0.0569
                                    0.3768
                                    0.1129
                                    0.0676
                                    0.0252
                                       
                                       
                                       
bpEF
(kg emit / kg total gas consumed)
                                       
                                       
                                       
                                       
                                       
                                       
                                       
                                     0.005
                                    0.0039
                                       
% of Total Emissions
percent
                                     48.73
                                     32.83
                                     9.72
                                     3.34
                                     0.00
                                     0.13
                                     0.00
                                     3.29
                                     1.95
                                       
Stack EF
mtCO2e Emitted/mtCO2e Consumed
                                       
                                       
                                       
                                       
                                       
                                       
                                       
                                       
                                       
                                    0.0936
EF Agreement
% Mar - Aug EF Diff.
                                      14
                                       4
                                       4
                                      21
                                       2
                                      18
                                       2
                                      28
                                      875
                                       2
The EPA is proposing certain provisions to minimize variability and uncertainty in EFs determined through stack testing: 
   1. The EPA is proposing provisions to ensure accurate measurements of gas consumption during the stack testing period (see Section 5.2). 
   2. The EPA is proposing that new stack EFs be developed from testing each reporting year, unless the variability in EFs meet certain variability criteria. Section 7 of this memo describes this in more detail. 
   3. As described in section 5.1.3 of this memo, the EPA is proposing provisions to ensure the representativeness of  EFs by requiring that there are no changes during the testing period in the centralized abatement systems (if present) or stack flow configuration.
6. Estimating Fab Emissions 
The proposed stack testing method would require facilities to estimate emissions from three kinds of F-GHG emissions: 1) emissions of intermittent low-use gases, 2) emissions from low-emitting stack systems, and 3) emissions from stack systems and gases tested. Methods the EPA is proposing for estimating each of these types of emissions are described below.
   1.  Emissions of intermittent low-use F-GHGs
The EPA is proposing that emissions of intermittent low-use F-GHGs be estimated using a modification of the 2006 IPCC Tier 2a method. Fabs would be required to use the same EFs as used to determine whether an F-GHG meets the intermittent low-use criteria (see Section 4 and discussion in the paragraph below on Tier 2a EFs), but would require the use of actual data for the current reporting year, including;
      * Fab abatement system uptime calculated as the complement of the downtime estimated per the methods described in Section 5.3;
      * Weighted-average fab-wide DREs, based on either default DRE values or actual measured DRE values;
      * Annual consumption of the intermittent low-use F-GHG, calculated per the provisions in subpart I based on a mass balance method, and;
      * The fraction of the intermittent low-use F-GHG used in tools with abatement systems as estimated by the fab.
The EPA is proposing that semiconductor manufacturers rely on new (since the 2006 IPCC Guidelines) Tier 2a EFs, which are based on emissions data supporting the default etch EF method (see Technical Support for Modifications to the Fluorinated Greenhouse Gas Emission Estimation Method Option for Semiconductor Facilities under Subpart I). For micro-electro-mechanical systems (MEMS) manufacturers, the EPA is proposing using the Tier 2a EFs for semiconductor manufacturing provided in the 2006 IPCC Guidelines because the manufacturing processes are very similar. The EPA is not currently updating the MEMS EFs based on any additional semiconductor data beyond what is in the 2006 IPCC Guidelines, because these devices are generally manufactured on older semiconductor manufacturing tools (i.e., 150 mm and 200 mm wafer sizes) and are just starting to move into manufacturing on newer tools (i.e., 300 mm). Without any additional data on emissions from MEMS processes, it was not feasible for the EPA to propose revised EFs for MEMS processes. The EPA is proposing that other electronics manufactures (liquid crystal displays (LCDs) and photovoltaic (PV) cell manufacturers) use Tier 2a EFs that are the same as those provided in the 2006 IPCC Guidelines. 
   2. Emissions from "low-emitting stack systems"
The EPA is proposing that emissions from low-emitting stack systems for reporting purposes be estimated using a Tier 2a method, again using the same EFs as used for determining whether a stack system is considered to be "low-emitting" and exempt from testing, and also using actual data from the current reporting year as described for intermittent low-use F-GHGs, above. However, when estimating emissions from low-emitting stack systems, a fab would also need to apportion the amount of F-GHGs consumed to low-emitting stack systems. The EPA is proposing that this be done using an apportioning model subject to the same apportioning model verification requirements as for apportioning gas consumption for other purposes in subpart I (e.g., apportioning gas usage to process types/sub-types and for the fraction used in tools with or without abatement systems). 
   3. Emissions based on a gas and fab-specific EF 
The third and final type of emissions that the EPA is proposing to require to be estimated for the stack test method are those that are based on the gas- and fab-specific EFs developed from stack tests using the methods being proposed by the EPA and described in this document. Using the calculation methods proposed by the EPA, estimating these types of emissions would require the same current reporting year data used for estimating emissions from low-emitting stack systems and intermittent low-use F-GHGs, with the exception that F-GHG consumption would be apportioned to the stack systems that are tested (in aggregate) and the gas- and fab-specific developed stack EFs.
7. Stack Testing Frequency
Table 7 below presents the proposed stack testing frequency requirements and the rationale for the proposed frequencies.
Table 7. Proposed Stack Testing Frequency Requirements and Supporting Rationale
                             Proposed Requirement
                                   Rationale
Initial Testing
Facilities would conduct three sets of tests of each stack system subject to testing, at a minimum of one test per year for three years. If they chose to do so, facilities can conduct more than one test per year, but each set of emissions testing would be separated by at least 2 months. 
The purpose of the initial testing is to determine the variability of stack EFs between tests. The variability in emissions based on the initial three tests would determine the frequency at which subsequent testing will need to be done. 
Subsequent Testing
After 3 sets of emissions testing are completed, if the relative standard deviation (RSD) of the CO2e EFs calculated from each of the 3 tests is <=15% and the RSD for all single F-GHG EFs that individually account for 5% or more of CO2e emissions is <= 20%, a reporter could use the arithmetic average of the three EF for calculating F-GHG emissions and defer testing for the next 4 years (unless a re-test is triggered). 
If the variability of the stack EFs is low, as indicated by the results of the RSD analysis, then testing would not need to be performed as frequently. However, if a particular facility has varying EFs due, for example, to changes in the type of products being manufactured, then testing would need to be more frequent.
The reporter would conduct an annual test, at a minimum, in the 5th year following the last stack test and repeat the RSD evaluation using the data from the three most recent tests. If the RSD criteria are not met, the reporter would continue testing annually at a minimum (but at least 2 months apart) and repeat the RSD analysis using the previous three sets of data. 
This provision has been included in the proposal to ensure that the EFs determined from stack testing are representative of current operations and emissions.
The reporter could do three tests in one year and try to meet the RSD criteria, but if they do not qualify to skip testing, they would still need to do annual testing beginning in the next year. If multiple sets of testing are done in one year, the average of those sets would be used for the annual report.
This provision is being proposed so that annual emission testing would still be required when the three initial tests are conducted in the first year and the results show that emissions are too variable to meet the criteria that would allow the facility to skip testing.
   
8. Changes at a Fab That Would Trigger Re-testing of Stack Systems
Data provided by SIA for various changes at a fab between 2009 and 2010 have demonstrated that changes in the mix of F-GHG and by-product precursors and in manufacturing processes can affect F-GHG emissions at the stack (see SIA, 2011). Because such changes would require that EFs developed from stack measurements would need to be kept up to date, the EPA, with input from SIA, developed and is proposing various criteria that would trigger re-testing of stack systems to revise EFs if the fab had previously qualified to skip annual testing. These criteria would not trigger a re-test if a facility is conducting annual testing. The proposed criteria for determining if a stack system would have to be re-tested are described below:
   * Consumption of a specific F-GHG used during the stack test changes by more than 10 percent of total annual gas consumption in CO2e, relative to gas consumption in CO2e for that gas during the year in which the most recent emissions tests were conducted. For example, if the use of a single F-GHG goes from 25 percent of total fab F-GHG consumption in CO2e to more than 35 percent of total fab F-GHG consumption in CO2e, that would trigger the need for a new test. 
   * A change in the use of an intermittent low-use gas that was not used during the emissions test and not reflected in the fab-specific EF, such that it no longer meets the proposed definition of an intermittent low-use gas (see Section 4). 
   *  A decrease by more than 10 percent in the fraction of tools with abatement systems, compared to the fraction of tools with abatement systems during the most recent emissions test. 
   * A change in the wafer size used by the fab since the most recent emissions test.
   * A change in a stack system that formerly met the criteria for being a low-emitting stack not subject to testing, such that it no longer meets those criteria. 
If any of these criteria are met, a fab would be required to re-test the stack systems and develop new F-GHG-specific and fab-specific EFs per the methods described in Section 5.
9. References
Intel (2010). Comments from Intel Corporation. Re: Mandatory Reporting of Greenhouse Gases: Additional Sources of Fluorinated GHGs: Proposed Rulemaking 75 Fed. Reg. 18652 (Apr. 12, 2010). Available in EPA docket EPA-HQ-OAR-2009-0927.
IPCC (2006). 2006 IPCC Guidelines for National Greenhouse Gas Inventories. The National Greenhouse Gas Inventories Programme, The Intergovernmental Panel on Climate Change, H.S. Eggleston, L. Buendia, K. Miwa, T Ngara, and K. Tanabe (eds.). Hayama, Kanagawa, Japan.  Available at: http://www.ipcc-nggip.iges.or.jp/public/2006gl/pdf/3_Volume3/V3_6_Ch6_Electronics_Industry.pdf. 
SIA (2011). SIA/ISMI Presentation on Stack Testing Alternative, November 30, 2011.  Available in EPA docket EPA-HQ-OAR-2011-0028.
SIA (2012a). Report to EPA on Etch Factor Proposal for Fab GHG Emissions Reporting, February 28, 2012.  Available in EPA docket EPA-HQ-OAR-2011-0028.
SIA (2012b). International SEMATECH Manufacturing Initiative Environmental Safety and Health Technology Center Etch Process Equipment Emissions Characterization Data (120228 etch report to EPA.xls), February 6, 2012. Available in EPA docket EPA-HQ-OAR-2011-0028.
SIA (2012c). Measured Gas Usage Rates During Stack Testing, April 28, 2012.  Available in EPA docket EPA-HQ-OAR-2011-0028.
SIA (2012d). Proposal for Dealing with Low Emissions Stack Systems that Will be Excluded from Testing Under the Stack Test Option, March 27, 2012.  Available in EPA docket EPA-HQ-OAR-2011-0028.
SIA (2012e). Response to EPA's Stack Test Questions, March 27, 2012.  Available in EPA docket EPA-HQ-OAR-2011-0028.
SIA (2012f). SIA Briefing Paper on Abatement Issues: Destruction Removal Efficiency (DRE) (Appendix 7 - 120110 SIA abatement briefing paper.pdf), January 10, 2012.  Available in EPA docket EPA-HQ-OAR-2011-0028.
SIA (2012g). Response to EPA's Stack Test Question 1. March 7, 2012.  Available in EPA docket EPA-HQ-OAR-2011-0028.
SIA (2012h). Intermittent Low/High Use Gas Criteria for Subpart I. April 27, 2012.  Available in EPA docket EPA-HQ-OAR-2011-0028.
U.S. EPA (2010).  Draft EFs for Refined Semiconductor Manufacturing Process Categories. United States Environmental Protection Agency, EPA-HQ-OAR-2009-0927-0073. May, 2010.

