Developed for Settlement Purposes (SIA v. EPA, D.C. Cir. No. 11-1024) Deliberative – Do Not Cite or Quote
Questions for SIA on Abatement System DRE, Uptime, and Related Requirements
On January 10, 2012, SIA provided data to EPA on abatement system uptime, abatement system inventories, and destruction or removal efficiency (DRE) measurement. This document represents EPA’s questions on the data and information submitted with respect to SIA’s concerns about DRE measurement. These questions are to ensure that the Agency understands SIA’s data/analyses and has enough information to support a new rulemaking to address these technical concerns. For the most part, the questions are clarifications on the information presented by SIA and not requests for additional data. We do not intend for these questions to result in extensive additional analysis or work, please let us know if that is not the case.
EPA appreciates the significant amount of information SIA has shared. It is very helpful and provides the largest abatement system DRE data set for Subpart I available to EPA. Based on the abatement system inventory information SIA has provided, EPA has estimated that the following installed percentages of systems by manufacturers:
· ~56 % of systems are Edwards,
· ~21% of systems are GST,
· ~14% of systems are CSK, and
· ~9 % of systems are from Unisem, AMAT and Ebara.
We have done some additional work to better define the current inventory of installed abatement units at US fabs. We have captured additional information from two device manufacturers to complete their inventories. We also determined that for the uninventoried units from the 2012 survey (approximately 900 abatement units at Mfr. C fabs), the distribution between combustion and plasma units is similar to the inventoried units for manufacturer C. This enables us to state that over 95% of the U.S. installed base of abatement units use combustion, and that over 70% of the combustion units are manufactured by Edwards.
Breakdown of POU F-Gas Abatement Devices Based on 2011 SurveyCompany / MFR / Combustion / Plasma / Total
A / GST / 277 / 277
A / CSK / 183 / 183
A / Unisem / 2 / 2
B / Edwards / 167 / 167
C / Edwards / 311 / 311
C / AMAT / 31 / 52 / 83
C / Ebara / 29 / 29
D / Edwards / 165 / 165
E / Edwards / 135 / 135
F / Edwards / 48 / 48
Total Units / 1348 / 52 / 1400
% of Total / 96.3% / 3.7%
% by Manufacturer
% Edwards / 59.0%
% GST / 19.8%
% CSK / 13.1%
% Unisem, AMAT, Ebara / 4.4% / 3.7%
We believe that the fact that the over 95% of the abatement units are combustion units enable the use of the available test data that we submitted to EPA to establish industry level DREs. Setting industry level, default DREs is justified by the following information:
o the tested units are representative of over 68% of the inventoried abatement units,
o combustion technology is straightforward; and
o the manufacturer’s design DREs are consistent across the different brands of combustion units.
Also based on the information SIA provided, EPA estimated that the data that SIA has collected was based on testing of abatement systems from predominantly one manufacturer, Edwards (~97% of tests).
Based on EPA’s initial review, it appears that the data in SIA’s DRE report supports the concept that achievable DREs vary by gas, process, and abatement system design/model. EPA agrees that an approach to provide a variety default DREs (e.g., by gas, process, abatement device, etc.) appears to be a viable option, with some further discussion and work.
Default DRE Measurements and Data
1. On pages 4 and 7, SIA indicates that in the testing, the abatement DRE test results for some CF4 and NF3 units were lower than expected and below the manufacturer’s DRE. SIA further indicates that one company “performed five retests after the abatement device was serviced to address a low DRE identified by the first test.” It is further stated that, while some units performed as intended, some operations and maintenance procedures needed to be revised to achieve the expected CF4 DREs.
a. Can SIA explain why the CF4 and NF3 DREs were low on the malfunctioning units?
A key contributing cause to the low DREs on abatement units treating CF4 is the fact that process flow rates (the volume into the abatement system) and/or the abatement discharge flow rates (the volume out) exceeded the manufacturer’s recommended flow rates. These combustion units have a recommended volume into the abatement unit of up to 250 standard liters per minute ( slpm) and a recommended volume out of up to 1,000 slpm. A review of the files “Etch Abatement Units with flow data. Xls”, which provides the flow rates in and out of the tested abatement units and the associated dilution rates, illustrate the relationship between these two flow rates and lower CF4 DREs. That relationship is not present for the DREs of other F-gases, which are more easily destroyed. The higher flow rates resulted either from inaccurately sized Mass Flow Controllers that were measuring/delivering the process gas and dilution flows to provide the correct mix of gases needed for safe handling of pyrophoric gases on the tool or the failure of the switching mechanism that adjusts the dilution rates based on the gas being fed to the process (for example, higher dilution rates are required for safety reasons during a H2 gas feed). For the CF4 DREs, the abatement system manufacturer is assessing other potential contributing causes of lower than expected DREs. SIA will provide an expanded response to this question sometime in March based on the additional information supplied by the equipment manufacturer.
For the CVD clean abatement units, which abate only NF3, there were only three tested units with a DRE lower than the manufacturer’s design DRE, out of the 35 units tested. Two systems were likely affected by high flow rate and one system was not. The file “CVD abatement units with flow rates.xls” provides the flow rates and dilution rates for the CVD clean abatement unit tests. The abatement system manufacturer is assessing other potential contributing causes which resulted in lower than expected DREs on the two CVD clean abatement units. SIA will provide an expanded response for both the NF3 and CF4 abatement units in March based on the additional information supplied by the equipment manufacturer.
b. Can SIA indicate whether there are any general criteria that could be applied to ascertain whether a unit is under performing or needs service without having to measure the abatement system DRE?
SIA is still considering the answer to this question and will provide a response before the end of March.
.
c. Can SIA clarify whether the initial testing data and the retest data for the retested units are in Tables VI and VII in Appendix B? If so, please identify this data.
Tables VI and VII have been updated to clearly identify the retests.
d. Can SIA indicate what criteria were used to determine whether a retest needed to occur?
Retests were done where abatement units were performing significantly below the abatement equipment manufacturer’s design DRE and where the initial test was done in the first set of tests in April and May so a retest could be scheduled during the second set of testing in August and September.
As discussed above, this round of abatement DRE testing was done to generate data for use in addressing DREs for the reporting rule. Where tested values are below the design DREs, they reduce the average DRE for the gas/process category and are represented in the average DRE in the installed base. Device manufacturers have an incentive to work to maximize the DRE of each unit as they are making significant investments in capital and operating expense for the units. The need for retesting should be left to the device manufacturer based on their corrective actions to improve the unit DREs, their current testing plans for the abatement units (where they are using a site specific DRE), and the expected increase in the DRE that will be realized through their investment in unit testing. Abatement unit testing in an operating semiconductor fab costs $6-8,000 per unit or $10-20,000 per unit when only a few units are being tested; moreover, testing requires production disruptions and engineering resources. On this basis, retesting should not be mandated in the rule so long as the data from the low DRE test is included in the site specific process/gas type average.
2. What method/protocol did SIA use for the DRE measurement tests (e.g., the EPA DRE Protocol) and did that method/protocol account for dilution by the abatement system?
The 57 tests performed at the one facility were performed in accordance with the procedure for Method 1 as described in the “Protocol for Measuring Destruction or Removal Efficiency (DRE) of Fluorinated Gas Abatement Equipment in Electronics Manufacturing March 2010” as specified in the Subpart I regulations. F-gases were flowed to the abatement units without operation of the tool plasma generator to simplify the testing process. Dilution rates were measured as required by the Protocol. Two example test reports, one for a tool specific abatement system and one for an area abatement system that abates several etch tools are provided in the files “FE05-002 DRE Test Report 13 Dec 2011.pdf” and “FK04 05 06 Backup DRE Test Report 13 Dec 2011.pdf” (This is the test of the area unit). These example files detail the methods use for measuring the dilution rates.
The majority of the testing reported from the other manufacturer was completed prior to March 2010. They were, however, done in general accordance with the Method 2 requirements and tested the DREs for the abatement units under actual fab operating conditions (with process gases flowing through the chamber with the tool plasma generator activated). Dilution measurements varied over the variety of tests due to various methods being developed over the course of time. Some of the dilution measurement methods used were primarily the QMS method specified in the rule and the post-POU FTIR dilution method. Variations of these methods may have been used, depending on the year in which the measurement was made.
In the context of the testing methodology, SIA is also asking EPA to include the 2009 ISMI Guideline tracer release/FTIR monitoring approach as an acceptable method for determining POU abatement system effluent flow. Inclusion of the effluent flow measurement method can be done by incorporating the method into the EPA Protocol referenced above or specifically adding the method as a referenced method in Section 98.7 of the Subpart I regulation. The efficacy and appropriateness of using the FTIR monitoring methodology has been demonstrated by the stack testing work completed by SIA members over the past 12 months. SIA has attached the report “Comparison of Fourier Transform Infrared (FTIR) and Quadrupole Mass Spectroscopy (QMS) Methods for Determining POU Abatement System Effluent Flow”. This report provides the data needed to justify the inclusion of this method in Subpart I. This method is more efficient, costs significantly less per test, and provides results which are comparable to the approved methods detailed in the “Protocol for Measuring Destruction or Removal Efficiency (DRE) of Fluorinated Gas Abatement Equipment in Electronics Manufacturing March 2010”.
In addition, the modifications to Subpart I should also set broad criteria, similar to that found in Subpart L 98.124(e)(1) & (2), that assures the quality of the testing procedure and analytical method used to perform required testing. 98.124(e) provides example methods, but also allows qualification of alternative methods, using accepted validation procedures, to enable development and early adoption of more efficient testing and analytical protocols using new or existing technologies.
3. Can SIA clarify how the measurements were taken at the facilities? Were they taken with different GHG flow rates over a variety of processes and tools? Can SIA provide the flow rates, or the normalized flow rates, of the gases for which DREs were measured?
The two abatement system test reports provided with this response detail the test procedures used for the 57 tests performed in 2011. The file “2011 DRE flow summary 02812.xls” details the range of flow rates used in each test for each gas.
The remaining tests were done using actual process conditions and flows. The table below is a summary of the range of gas flows and the pre- and post-dilutions for the process conditions tested for each category and gas type.
# of emissions tests / NF3 Range (sccm) / SF6 Range (sccm) / CF4 Range (sccm) / CHF3 Range (sccm) / POU Inlet flow range (L/min) / POU Outlet flow range (L/min)in-situ plasma chamber clean / 6 / 1000-3000 / 100-200 / 750-1000
Plasma Etch / 2 / 10-30 / 100-300 / 70-200 / 20-80 / 300-600
remote plasma chamber clean / 3 / 200-2000 / 50-200 / 400-800
4. Can SIA please clarify whether all measurements were done on abatement systems specifically designed to abate F-GHGs?
All tested abatement systems were specifically designed to treat F-gases.
5. Can SIA please indicate if their testing supports any conclusions regarding how the performance of abatement systems changes (or does not change) between maintenance events?