                                       
Subject: EPA/SIA Phone call
Date: October 24, 2012, 12-1 PM
Re: Questions regarding the EPA proposal, Greenhouse Gas Reporting Program: Proposed Amendments and Confidentiality Determinations for Subpart I  
Background:  This phone call between EPA and the Semiconductor Industry Association (SIA) was held at SIA's request following SIA's initial review of the EPA proposal, Greenhouse Gas Reporting Program: Proposed Amendments and Confidentiality Determinations for Subpart I, [77 FR 63538, October 16, 2012].  SIA provided the questions below to EPA via e-mail on October 8, 2012 (following the pre-publication version of the proposal being posted on the Greenhouse Gas Reporting Program (GHGRP) website on August 31, 2012) and also verbally reviewed the questions on the phone call.  EPA provided responses to SIA's questions verbally on the phone call and is documenting these responses for docket number EPA-HQ-OAR-2011-0028 below.  This memorandum also contains additional response information not provided during the phone call.
Attendees:
EPA  -  Alexis McKittrick, Anhar Karimjee, Deborah Ottinger, Jennifer Bohman, Kirsten Cappel
SIA - David Isaacs
SIA Consultants - John Holmes, Robert Crawford, Kevin Fay
Fairchild Semiconductor - Dave Lancaster
IBM - Dave Speed, Jay Dietrich
Intel - Tim Higgs, Melissa Gresham
Texas Instruments - Tina Gilliland, Jack Chang
Samsung - Mike Sherer (consultant)
ISMI - Laurie Beau
Eastern Research Group  -  Brian Palmer, Philip Crawford
SIA Question 1: We are wondering how the revised Tier 2a factors that appear in Tables I-11 and I-12 [of the proposed rule amendments] were derived, as we are unable to duplicate them for C2F6 or NF3, when using the SIA gas use distributions and the updated default factors for each process and subprocess.
EPA Response 1:  From page 22 of 25 in the memorandum available in the docket titled Technical Support for the Stack Testing Option for Estimating Fluorinated Greenhouse F-GHG Emissions from Electronics Manufacturing Facilities under Subpart I: "The EPA is proposing that semiconductor manufacturers rely on new (since the 2006 IPCC Guidelines) Tier 2a EFs [emission factors], 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)." 
The referenced memo (Technical Support for Modifications...), discusses default EF revisions, along with what models were considered and used, but focuses on the more complex EF development (i.e. Tier 2c).  Tier 2a EFs are not specifically mentioned, largely because the method for developing revised default Tier 2a EFs is similar to that used for the other, more complex default EF categories.  To develop the Tier 2a EFs, EPA generally calculated the simple arithmetic average across all the EF measurements available for each gas and wafer size across all process types. SIA indicated via phone that they were using an average weighted by gas distribution, which could explain some of the discrepancy. 
 For NF3 specifically, there are two other factors that may explain any differences between SIA's results and EPA's results.  First, EPA combined data for the NF3 remote clean process type with the data for the other NF3 process types.  This is different from the approach that was taken in the 2006 IPCC Guidelines (the source for the original Tier 2a factors), which presented separate default EFs for NF3 remote clean and for the other NF3 process types.  Second, in combining the NF3 remote clean process type with the other NF3 process types, EPA inadvertently used only one value for the NF3 remote clean process type for each wafer size.  Specifically, EPA used the average EF value for that wafer size.  As a result, the NF3 remote clean process type was under-weighted in the combined default NF3 EFs.       
For C2F6, it is unclear why SIA is having difficulty replicating EPA's proposed EF.  The data point counts for both EPA and SIA appear to match.  EPA and SIA were unable to determine via phone where the inconsistency lies, but SIA indicated that the difference was small enough to not warrant further discussion prior to the comment period closing for this proposal.  SIA did clarify that this discrepancy exists for the C2F6 200mm EF, but not for the corresponding 300mm EF.
To assist with resolving these discrepancies, SIA requested a spreadsheet detailing the NF3 and C2F6 Tier 2a EF calculations. The spreadsheet posted concurrently with this memorandum, titled "Tier 2a NF3 and C2F6 Emission Factor Calculations", shows the EF calculations and their results for the NF3 and C2F6 default EFs (1) as presented in the proposed rule, (2) when NF3 remote clean is treated as a separate process type (consistent with the original Tier 2a defaults in the 2006 IPCC Guidelines), and (3) when NF3 remote clean is combined with the other NF3 process types using all of the NF3 remote clean data (except the outliers).  For scenarios (2) and (3), both simple and weighted averages are presented, resulting in a total of five sets of Tier 2a default EF values.  For convenience, the spreadsheet also shows the four outlier data points that were removed from the NF3 EF calculation.  (Note that these four outliers were also identified in the 2010 Technical Support Document on page 36 in Appendix B.)
SIA Question 2: We do not understand the basis of EPA's assessment of the SIA case for the size of apportioning errors that appears on pp. 10-13 and in Table 2 (p. 14) of the document titled: Technical Support for Modifications to the Fluorinated Greenhouse Gas Emission Estimation Method Option for Semiconductor Facilities under Subpart I. In this part of the document it says that "a range of apportioning uncertainties estimated by SIA, varying between 56 and approximately 240 percent (SIA, 2011c and SIA, 2012a)". How were the SIA input values translated into gas apportioning errors and used in the Monte Carlo analysis?
EPA Response 2: The detailed analysis that was done and the inputs that were used in the EPA analysis shown in the docketed memo titled Technical Support for Modifications to the Fluorinated Greenhouse Gas Emission Estimation Method Option for Semiconductor Facilities under Subpart I reflect information from SIA submitted on gas apportioning uncertainties (Report to EPA on Etch Factor Proposal for Fab GHG Emissions Reporting, dated February 28, 2012).  The values were adjusted according to the method described on pages 24 and 25 of the referenced report.  SIA also asked during this call whether uncertainties related to chamber cleaning were included in EPA's analysis.  EPA confirms that these uncertainties were included.
SIA Question 3: SIA members are collecting additional emissions factors data and DRE [destruction or removal efficiency] data that can be used to update the default values for each. As in the past, SIA intends to provide both the test results and an update of the default values when they are integrated into the current data base. The notice includes a template for data submittal. SIA had developed and has been using its own template. Two points in that regard. First, we are not gathering all of the data in the EPA template. We can give EPA a list of those not being collected and our reasoning. Second, will it be acceptable to submit the data in our template or would you need us to put the data into your template?
EPA Response 3: From page 63552 of the October 16, 2012 Federal Register notice: "Commenters are encouraged to submit available data with their comments using the `Electronics Manufacturing Data Request Sheet' (see Docket ID No. EPA - HQ - OAR - 2011-0028). Commenters can fill out the `Electronics Manufacturing Data Request Sheet' and submit the data to Docket ID No. EPA - HQ - OAR - 2011 - 0028 for consideration by the EPA on whether to update the proposed default emission factors for semiconductor manufacturing. If the EPA does update the proposed default emission factors using such new data, if approved by the EPA, for the final rule, it will do so using the same methodologies as described in the `Technical Support for Modifications to the Fluorinated Greenhouse Gas Emission Estimation Method Option for Semiconductor Facilities under Subpart I,' Docket ID No. EPA - HQ - OAR - 2011 - 0028). The EPA will use the same criteria for accepting new data that were used in accepting data as specified in that document."  A similar statement is also provided in the proposal preamble that references the DRE support material (77 FR 63564).
As stated above, commenters are encouraged to submit data using the referenced EPA data request sheet.  SIA indicated via phone that several pieces of data may be missing from DRE data they plan to submit with their comments, including the DRE relative standard deviation, the dilution factor, and whether or not the EPA DRE Protocol benchmark is met.  EPA notes that, if the EPA DRE Protocol is followed, commenters submitting data should have enough information to complete all of the fields in EPA's data sheet template, including those related to dilution factor and standard deviations. EPA stresses that, as stated above, data submitted must supply EPA with enough information to assure data quality and to use the same methodologies that are described by the technical support documents for this proposal. All data are subject to EPA review and may not be incorporated in the calculations for subpart I default EFs and default DREs if sufficient information is not provided. Regardless of the template that is ultimately used, SIA is encouraged to submit its own analysis of the data provided, as long as such additional information does not in any way limit the amount of raw data submitted to EPA.
SIA Question 4: We see that EPA has added the results of 11 DRE tests to the record. Please provide SIA with the data records (test identifications and DRE results) for the extra 11 etch/CF4 tests from 2010 testing that have been rolled into the default DRE analysis.
EPA Response 4: The original November 2010 technical support document (TSD) presented 11 DRE measurement results for etch/CF4. These appear to be the 11 DRE tests to which SIA is referring. Please see Table 3-3 on page 21 of the November 2010 TSD.  Note that these results are referenced on page 3 of 14 of the docketed memorandum titled, Technical Support for Accounting for Destruction or Removal Efficiency for Electronics Manufacturing Facilities under Subpart I: "Table 1 presents a summary and analysis of the available etch DRE data SIA submitted to the EPA and DRE data available through the EPA DRE measurement studies (SIA, 2012a and November 2010 TSD)". 
As referenced in Table 3-3, the original November 2010 TSD used data from Table IV (8 data points), Table VIII (2 data points), and Table IX (1 data point) of Developing a Reliable Fluorinated Greenhouse Gas (F-GHG) Destruction or Removal Efficiency (DRE) Measurement Method for Electronics Manufacturing: A Cooperative Evaluation with IBM (EPA 430-R-10-004), Office of Air and Radiation Office of Atmospheric Programs, Climate Change Division, U.S. Environmental Protection Agency, Washington, DC. Available at: http://www.epa.gov/highgwp/semiconductor-pfc/documents/ibm_report.pdf.
SIA Question 5: In the preamble at page 155 of the prepublication pdf [p. 63569 of the October 16 FRN], EPA references statistical methods to validate reported data elements. SIA members would like to know more - what methods and how would they be applied?
EPA Response 5: This is a reference to EPA's verification process for GHG data submitted via the electronic Greenhouse Gas Reporting Tool (e-GGRT) system.  EPA has a multi-step data verification process. First, the EPA data entry tool used by facilities conducts data checks during the data entry phase. This occurs even before the data are submitted to EPA. Second, once the data have been submitted, EPA conducts a number of automated data checks that include ensuring that reports are internally consistent, checking the data against expected ranges for similar facilities and industries, and statistical analysis. Based on the results of the automated checks, EPA conducts a staff review of the reported data. EPA then follows up with facilities to understand any inconsistencies or unexpected results, and if necessary, to resolve any mistakes that may have occurred.  This process occurs for all data submitted through the e-GGRT system (not for subpart I only).
SIA Question 6: The stack test method, as we read it, requires certification of abatement devices and a maintenance plan or the stack method cannot be used. Is the way we are reading it the way it is intended?
EPA Response 6: SIA's reading of this proposed requirement is correct. If a reporter were using the proposed stack testing method, they would be required to follow the quality assurance (QA) procedures in proposed 40 CFR 98.94(f) for abatement systems. These include certification and documentation that the abatement systems are properly installed, operated, and maintained according to manufacturers' specifications; and preparation of a site maintenance plan for abatement systems that is maintained in the facility's records. (See p. 63584 of the October 16 FRN for the amendatory text for 40 CFR 98.94(f).)  As noted on pages 63562 and 63564 of the preamble to the proposed rule, these procedures are intended to ensure that DRE measurements are representative of abatement system performance throughout the year, and vice versa. SIA requested clarification during the call regarding why this requirement is being proposed for the stack testing method.  Although stack tests do not separately measure DREs, they do reflect abatement system performance during the test.  EPA's proposal to apply the QA procedures in 40 CFR 98.94(f) to fabs using the stack testing method would ensure that abatement systems are maintained at the same level throughout the year, thus further ensuring that the measurements during the stack test are representative of annual emissions. 
