»EPA
United States
Environmental Protection
Agency
SFe Emission Reduction
Partnership for the Magnesium Industry
Characterization of Cover Gas and
Byproduct Emissions from
Secondary Magnesium
Ingot Casting
Office of Air and Radiation
Office of Atmospheric Programs, Climate Change Division
EPA 430-R-08-008
-------
Characterization of Cover Gas and Byproduct Emissions from
Secondary Magnesium Ingot Casting
Scott C. Bartos
U.S. Environmental Protection Agency
Climate Change Division
1200 Pennsylvania Avenue, NW (6202J)
Washington, DC 20460
October 2008
EPA 430-R-08-008
-------
Table of Contents
Page
Acknowledgement vi
Executive Summary ES-1
1.0 Introduction 1-1
2.0 Methodology 2-1
2.1 Principles of FTIR Monitoring 2-1
2.2 Principles of RGA Monitoring 2-5
2.3 Ambient Air Dilution Considerations 2-7
3.0 Monitoring Results 3-1
3.1 Casting Hood Monitoring 3-1
3.2 Worker Exposure Monitoring 3-6
4.0 Cover Gas Destruction 4-1
4.1 Determining Dilution 4-1
4.2 Determining Cover Gas Destruction 4-6
5.0 Discussion 5-1
5.1 Cover Gas Test Observations 5-1
5.2 Climate Change Potential Discussion 5-3
5.3 Uncertainty Discussion 5-4
Appendix A. Calibration and Diagnostic Checks A-l
Appendix B. Measurement Study Protocol B-l
in
-------
List of Tables
Page
ES-1 Cover Gas Average Concentrations and Observed Destruction ES-3
ES-2 Global Warming Potential of Alternative Cover Gas Mixtures ES-4
1-1 Test Schedule and Process Conditions 1-2
1-2 Magnesium Ingot Casting Machine Parameters 1-2
2-1 Parameters for Maj or Contaminants and Spectroscopic Interferants 2-2
2-2 Relative Isotopic Abundance for Argon and Krypton 2-6
3-1 Data Summary for SF6/CO2 Cover Gas Mixture 3-2
3-2 Data Summary for MTG-Shield™ using Novec™ 612 3-3
3-3 Data Summary for MTG-Shield™ using Novec™ 612 3-4
3-4 Data Summary for SC>2 Cover Gas Mixture 3-5
3-5 Worker Exposure Monitoring 3-6
4-1 Dilution Percentages (DP) Calculated by Kr Tracer and Direct Ar Monitoring 4-5
4-2 Dilution Percentages (DP) Calculated from Noncasting and Direct Concentrations .... 4-6
4-3 Percent Destruction for Cover Gas Testing 4-7
5-1 Comparison of 100-year GWP Estimates from the Intergovernmental Panel on
Climate Change (IPCC) Fourth (2007) Assessment Report 5-3
5-2 Normalized GWP Comparison of Measured Emissions from Inside the Casting
Hood 5-5
5-3 GWP (Weighted by Flow Rate) Comparison of Measured Emissions from Inside the
Casting Hood 5-6
IV
-------
List of Figures
Page
2-1 Casting Hood and Sampling System Schematic 2-4
2-2 RGA Component Block Diagram 2-5
4-1 RGA Dilution Measurements, 11 September 2007 4-2
4-2 RGA Dilution Measurements, 12 September 2007 4-3
4-3 RGA Dilution Measurements, 13 September 2007 4-4
A-l Calibrated RGA Response for Argon A-4
A-2 Calibrated RGA Response for Krypton A-5
-------
Acknowledgements
The analytical measurements, data interpretation, and report preparations were funded by
the U.S. Environmental Protection Agency under contract GS-10F-0124J to ICF International.
The authors wish to express their appreciation and thanks to Magnesium Refining Technologies
(MagReTech) and staff, especially Ken Balser, for contributing not only their facilities but also
their valuable assistance and advice to this measurement study. The support of Matheson Tri-
Gas, Polycontrols Inc., and 3M™ for providing the cover gases, their expertise, and trial staff for
this study is also gratefully acknowledged.
VI
-------
Executive Summary
This measurement study was conducted to evaluate the greenhouse gas (GHG) emissions
and occupational exposure associated with three cover gas technologies used in a magnesium
alloy ingot casting machine. Sulfur hexafluoride (SF6) is widely used for the protection from
oxidation of molten magnesium; but with the goal of eliminating the use of SF6 in this
application by 2010, the magnesium industry and U.S. Environmental Protection Agency (EPA)
have been evaluating the use of alternative gases. This study expands upon previous research by
continuously monitoring multiple sample points and by testing cover gas mixtures in an ingot
casting hood environment - previous research used single sampling points and examined die
casting holding furnaces.l This study examined the use of SF6, pentafluoroethylhepafluoro-
isopropylketone (Novec™ 612), and sulfur dioxide (SO2) on an ingot casting machine located at
a Magnesium Refining Technologies (MagReTech) facility located in Bellevue, Ohio. Process
and machine operating parameters were maintained at similar levels when each cover gas
mixture was evaluated. Each cover gas mixture was injected into the ingot casting hood and
sampled from the hot and cold zones of the casting hood to characterize emissions and
byproducts as the cover gases interact with the magnesium melt surface and undergo thermo-
degradation. The results reported are from measurements taken from multiple points inside the
casting hood, and from an ambient air sampling point in the casting hood operator room. Results
are presented for four sample points: the upstream sampling point in the hot zone (hu), the
downstream sampling point in the hot zone (hd), the upstream sampling point in the cold zone
(cu), and the downstream sampling point in the cold zone (cd). Table ES-1 summarizes some of
the details and results from the ingot casting hood component of the study. Measurements were
conducted using slightly varied mixtures of each cover gas. The cover gas destruction rates
listed in Table ES-1 have been corrected for dilution effects.2
SF6/SO2 with CDA/CO2 Carrier Gas
The only destruction byproduct measured while using SF6/SO2 with clean dry air
(CDA)/CC>2 as a cover gas was FTP, with concentrations ranging from 0.02 to 0.37 ppmv over
three different testing periods. Other byproducts, including COS, €82, H2S, and F^SO/t, (see
Table 2-1 for a listing of chemical formulas and compound names) were not analytically
detected. CH4 was observed at average concentrations of 2.6 to 82.8 ppmv during the three tests;
the normal ambient air concentration of CFLi is ~2 ppmv. The high concentration of CFLt is most
likely due to the periodic use of to natural gas burners to preheat the ingot molds, therefore CH4
1 US EPA. Characterization of Emissions and Occupational Exposure Associated with Five Cover Gas
Technologies for Magnesium Die Casting, EPA 430-R-07-008, August 2007.
2 The term destruction is utilized throughout the remainder of this report to represent the thermo-degradation and
disassociation of the cover gas agent resulting in byproduct formation and melt protection.
ES-1
-------
measured in the ingot casting hood was assumed to originate from fugitive emissions in the
facility rather than process emissions associated with cover gas application.
MTG-Shield™ using Novec™ 612 with CDA/CO2 Carrier Gas
The only destruction byproducts measured while running MTG-Shield™ using Novec™
612 with CDA/CO2 as a cover gas were HF and CHF3. COF2, CF4, C2F6, C3F8, OF2, CH2O2, and
H2CO were expected byproducts that were not analytically detected. HF was detected at
concentrations ranging from 0.05 to 1.5 ppmv for the three Novec™ 612 testing periods. CHFs
was twice detected at the downstream sampling point in the hot zone at concentrations ranging
from 0.15 to 0.30 ppmv. Perfluoroisobutylene (PFIB), an occupational hazard and primary
byproduct of concern, was monitored for, but not detected during this study.
SO2 with CDA Carrier Gas
There were no destruction byproducts detected that could be attributed to the SO2 with
CDA cover gas mixture. The only detectable compounds were ambient air and combustion-
related compounds (H2O, CO, CO2, and CH4). Methane was detected at average concentrations
ranging from 3.2 to 3.6 ppmv. Sulfuric acid and H2S were not measured at concentrations above
their minimum detection limits within the ingot casting hood.
Observed Percent Destruction for Cover Gases
Table ES-1 lists the destruction estimates for all cover gases examined. The destruction
estimates, which are corrected for dilution effects (i.e., the effects of air ingression into the ingot
casting hood), are calculated as the percent difference between the expected dilution corrected
delivery concentration and the measured concentration in the casting area. Average destruction
estimates for Novec™ 612 and SF6 were on the order of four percent and three percent,
respectively. In comparison, destruction estimates for SO2 were on the order of 39 percent for
this study. It should be noted that high levels of dilution found in this study and associated
uncertainty resulted in calculated destruction rates for some tests being unreasonable (i.e.,
negative) and these values were treated as zero destruction results.
The destruction rates estimated for SF6 in this study were significantly lower than what
was estimated during previous research evaluating die casting holding furnaces (on the order of
20 to 30 percent). This is likely due to the much higher levels of dilution and reduced thermo-
chemical intensity of the casting hood environment.
ES-2
-------
Table ES-1. Cover Gas Average Concentrations and Observed Destruction
Test#
(site)
Ihu
leu
2hu
2cu
Shu
3cu
3hd
3cd
4hu
4cu
5hd
5cd
6hu
6cu
6hd
6cd
Thu
7cu
8hd
8cd
9hu
9cu
Cover Gas Mixture
Components
SF6 / SO2 / CDA / CO2
SF6 / SO2 / CDA / CO2
SF6 / SO2 / CDA / CO2
SF6 / SO2 / CDA / CO2
SF6 / S02 / CDA / C02
SF6 / S02 / CDA / C02
SF6 / S02 / CDA / C02
SF6 / S02 / CDA / C02
Novec™612/CDA/C02
Novec™ 612 / CDA / CO2
Novec™612/CDA/C02
Novec™ 6127 CDA / CO2
Novec™ 6 127 CDA /CO2
Novec™ 6127 CDA / CO2
Novec™ 6 127 CDA /CO2
Novec™ 612 / CDA / CO2
Novec™ 6 127 CDA /CO2
Novec™ 612 / CDA / CO2
SO2 / CDA
S02 / CDA
S02 / CDA
S02 / CDA
Time
1730-1930
1730-1930
1115-1345
1115-1345
0915-1200
0915-1200
0915-1200
0915-1200
1500-1709
1500-1709
1710-1730
1710-1730
0915-1140
0915-1140
0915-1140
0915-1140
1420-1620
1420-1620
1340-1540
1340-1540
0740-1000
0740-1000
Cover Gas
Mixture
Flow3
(1pm)
270
135
198
99
270
135
270
135
189
95
189
95
191
96
191
96
189
95
191
95
191
96
Cover Gas
Delivery
Conc.b
(ppmv)
10,675
10,675
7,554
7,554
8,463
8,463
8,463
8,463
2,000
2,000
1,500
1,500
1,102
1,102
1,102
1,102
1,663
1,663
10,000
10,000
12,000
12,000
Cover Gas
Dil. Corr.
Conc.c
(ppmv)
1,459
624
1,285
490
954
479
301
212
220
76
40
30
124
62
39
28
125
241
356
250
1353
679
Cover Gas
Measured
Cone.
(ppmv)
1,407
572
1,251
456
1,232
640
424
333
219
77
47
27
96
161
37
74
131
228
111
128
946
296
Cover Gas
Destruction
Factor
(percent)
4%
8%
3%
7%
-0%
-0%
-0%
-0%
-0%
-0%
-0%
10%
23%
-0%
4%
-0%
-0%
5%
22%
49%
30%
56%
a Approximate, estimated by reading flow rates on gas delivery manifold rotameters (uncalibrated). Two-thirds of the flow went to
the hot zone and one third to the cold zone. Additional details on cover gas mixture can be found in Table 1-1.
b Measured directly at manifold; only for primary gases of concern (SF6, Novec™ 612, and SO2) for the three cover gas systems.
0 Dilution corrected concentration based on dilution estimates in Chapter 4. These are the expected concentrations with no
destruction occurring.
Occupational Exposure Monitoring
Since each cover gas used in this study can result in emissions that may be harmful to
exposed workers, monitoring of the ambient air in the casting hood worker room was performed
using another FTIR. This component of the study was especially relevant to SC>2 as well as the
byproduct HF, due to their stringent occupational exposure limits. The breathing zone located
near the worker responsible for skimming ingot surfaces was continuously monitored during the
study. Table 3.5 lists the results for the occupational exposure monitoring performed during this
study. On average, the only compound found in the worker room above detectable levels was
SF6, which was only present when the SF6/SO2 cover gas was being used. Sulfur dioxide,
ES-2
-------
Novec™ 612, and HF were not present in the worker room above detectable levels for any of the
cover gas mixtures and SFe was not detected during the Novec™ 612 or SC>2 tests.3
Potential Climate Impact
A key factor in evaluating alternative cover gas compounds was their composite global
warming potentials (GWPs) as compared to SFe. Global warming potentials are based on the
heat-absorbing capability and atmospheric lifetime of each gas relative to that of carbon dioxide.
The GWP provides a construct for converting emissions of various gases into a common
measure, denominated in carbon dioxide equivalents. For each cover gas compound and its
applicable destruction byproducts, a composite global warming impact estimate was developed
using the IPCC Fourth Assessment Report (AR4) GWP values.4 The overall GWP-weighted gas
emission rate for each cover gas regime was estimated using the measured average
concentrations of each gas, their molecular weights, and the delivery cover gas flow rates. This
resulted in a normalized CC>2 emission equivalent for each alternative cover gas that could be
directly compared to the CC>2 emission equivalent of the existing SF6/SO2 system.
Based on this approach, results indicate that both the Novec™ 612 cover gas mixture and
the SC>2 cover gas mixture have a GHG emission impact - weighted by cover gas flow - that is at
least 99 percent lower than SF6. Table ES-2 presents the results of the global warming potential
analysis for the alternative cover gas mixtures examined in this experiment.
Table ES-2. Global Warming Potential of Alternative Cover Gas Mixtures
Cover Gas Mixture
Novec™ 612 / CD A / CO2
SO2 / CDA
GHG Emissions Relative to
Existing SF6/ SO2 system
(percent reduction)
>99%
>99%
3 All applicable safety precautions (e.g., operational procedures) should be followed when using SO2.
4 IPCC, Climate Change 2007: The Scientific Basis. Intergovernmental Panel on Climate Change, 2007, Cambridge
University Press. Cambridge, U.K.
ES-4
-------
1. Introduction
This report presents and interprets the results of a series of emission measurements taken
of air-entrained cover gas mixtures at a single magnesium alloy ingot casting machine. The
measurements were conducted by URS Corporation (URS) at the Magnesium Refining
Technologies (MagReTech) facility located in Bellevue, Ohio, throughout the week of 9
September 2007. Emissions were extracted and continuously analyzed in near real-time with
Fourier Transform Infrared (FTIR) spectroscopy and quadrupole mass spectrometry Residual
Gas Analysis (RGA). These analysis techniques enabled the simultaneous quantification of
multiple concentrations in the cover gas environments at sub-ppmv-level sensitivities.
The three base gases evaluated in this study were sulfur hexafluoride (SF6),
pentafluoroethylhepafluoroisopropylketone (known by trade name Novec™ 612), and sulfur
dioxide (SO2). The cover gas mixtures are used to protect molten magnesium against surface
burning during ingot casting. The primary objectives of this study are listed below.
• Characterize the cover gas destruction at this particular ingot casting tool. Destruction
rates impact overall greenhouse gas (GHG) emissions. Destruction is defined as the
percentage of base cover gas consumed by the process, whether by breakdown to a
magnesium fluoride (MgF2) film and subsequent chemical byproducts, or by direct
conversion to byproducts from the thermal conditions and chemistries residing in the
casting space environment.
• Characterize the ambient air dilution into the ingot casting and cooling environments.
The hot zone and cold zone sections of the casting hood are not sealed, so air intrusion
was expected to be significant. Ambient air dilution must be factored into the cover gas
consumption considerations so that destruction rates can be properly filtered out from
overall concentration reductions.
• Characterize the chemical byproducts produced for each cover gas mixture during ingot
casting. The types and relative amounts of byproducts vary greatly depending on the
base cover gas and the concentration at which it is used. The chemical byproducts
produced also influence the global warming potential of the cover gas mixtures.
• Identify and detect low concentration occupational exposure emissions for each cover gas
mixture. Using the most sensitive FTIR system available, monitor a worker area in close
proximity to the hot zone for base cover gas and byproduct emissions.
The measurement schedule and test conditions are summarized in Table 1-1. Rather than
being fixed, these conditions were what was encountered during facility operations at the time of
1-1
-------
testing and resulted in a variety of cover gas mixture compositions over different alloy castings.
The ingot casting machine parameters are summarized in Table 1-2.
Table 1-1. Test Schedule and Process Conditions
Date
(m/dd/yy)
9/10/07
9/11/07
9/13/07
9/11/07
9/11/07
9/12/07
9/12/07
9/13/07
9/14/07
Approx.
Casting Time
(Local Time)
1730-1930
1115-1345
0915 - 1200
1500-1709
1709 - 1730
0915-1140
1420 - 1620
1340-1540
0740 - 1000
Cover Gas Mixture
Components3
SF6 / SO2 / CDA / CO2
SF6 / SO2 / CDA / CO2
SF6 / SO2 / CDA / CO2
Novec-612/CDA/C02
Novec-612/CDA/CO2
Novec-612/CDA/CO2
Novec-612/CDA/CO2
SO2 / CDA
SO2 / CDA
Cover Gas Mixture
Flows"
(scfm)
0.08/0.02/6.8/7.4
0.05/0.03/4.2/6.2
0.07/0.04/5.5/8.7
0.015/2/8
0.015/2/8
0.0125/2/8
0.015/2/8
0.10/4/6
0.12/4/6
Cover Gas
Delivery
Conc.c
(ppmv)
10,675
7,554
8,463
2,000
1,500
1,102
1,663
10,000
12,000
Alloy Type
AM50
AZ91D
AZ81
AZ91D
AZ91D
AZ91D
AZ91D
AZ91D
AZ91D
CDA = compressed dry air
b Approximate, estimated by reading flow rates on gas delivery manifold rotameters
0 Measured directly at cover gas mix
(uncalibrated)
Table 1-2. Magnesium Ingot Casting Machine Parameters
Parameter
Facility
Ingot Casting Machine Type
Ingot Weight (Ibs)
Holding Furnace Capacity (Ibs)
Alloy Type
Ingot Casting Rate (seconds/ingot)
Mg Pump Type
Metal Throughput (Ibs/hr)
Heat Casting Duration (hours)
Ingot Mold Temperature (°F)
Ingot Residence Time - Hot Zone (min)
Ingot Residence Time - Cold Zone (min)
Ingot Pour Control
Machine Specification3
MagReTech: Bellevue, OH
Belt Caster
25
16,000 (per heat)
All
12
Centrifugal Pump
=5,000
3-4
120-200
0.5
2
Automatic Feed w/ Operator Override
aAs provided by MagReTech
1-2
-------
2. Methodology
All gas samples, whether from the casting hood environments or the worker exposure
area, were extracted continuously from single points in space. The four sampling points located
in the ingot casting area were 1) the upstream sampling point in the hot zone (denoted "hu"), 2)
the downstream sampling point in the hot zone (denoted "hd"), 3) the upstream sampling point in
the cold zone (denoted "cu"), and 4) the downstream sampling point in the cold zone (denoted
"cd"). The sampling system is described in more detail in Section 2.1.2 and the schematic is
presented as Figure 2-1.
This section of the report describes the field analytical methods used to probe the gas
samples (FTIR and RGA are explained in Sections 2.1 and 2.2, respectively) and how ambient
air dilution was determined within the casting hoods (Section 2.3).
2.1. Principles of FTIR Monitoring
Almost every chemical compound absorbs some amount of infrared (IR) light in a
particular region of the mid-IR spectrum. These absorption properties can be used to identify
and quantify chemical compounds in a complex mixture of gases. Beer's Law states that the
magnitude of IR absorbance by a compound is directly proportional to the product of its
concentration in the mixture and the sample cell optical path length. This product is otherwise
known as the optical depth of a compound. The extractive FTIR instruments used by URS are
able to achieve ppb detection levels because the optical path length within the measurement cell
is magnified many times by reflecting the IR beam between a series of mirrors before it reaches
the detector. The mirrors provide a fixed optical path length best suited to the gas mixture being
sampled. For this study, optical path lengths of 10 m (for worker exposure monitoring), 5.11m
(for hot zone casting head space monitoring), and 20.1 m (for cold zone casting head space
monitoring) were utilized.
2.1.1. The FTIR Spectrum Analysis Method
IR spectrum analysis matches the features of an observed spectrum to those of reference
gases of known concentration. If more than one feature is present in the same region, then a
linear combination of references is used to match the compound feature. The standards are
scaled to match the observed band intensities in the sample. This scaling also matches the
unknown concentrations. An IR spectrum can be collected and analyzed in approximately one
second, but spectra are normally averaged over 1- to 5-min integration periods to produce
adequate signal-to-noise limits and ppb detection levels.
The scaled references are added together to produce a composite that represents the best
match with the sample. A classical least squares mathematical function is used to match the
2-1
-------
absorption profiles of the standards with those of the observed spectrum in specified spectral
analysis regions. The compounds of interest and the compounds expected to cause spectral
interference are included in the analysis region.
The spectrum analysis methods used for this study were developed by selecting the
spectral regions that were least affected by primary IR absorbers (in this case, H^O and CO2)
while also producing the best detection limits possible for cover gas compounds and potential
byproducts. Target compounds were determined prior to sampling based on previous tests of
similar cover gas composition. The analysis methods were iteratively refined by analyzing
representative sets of IR spectra while varying quantitative analysis parameters until optimum
methods were established. Methods are optimum when the 95 percent confidence levels (the
errors indicating goodness-of-fit) and the absolute bias of all analytes are minimized. Table 2-1
lists the calibration reference set ranges, as they pertain to their respective cover gas mixture, for
all the compounds monitored throughout the MagReTech study. This represents a complete list
of potential and existing contaminants, though not all of these contaminants were observed
during the study, as later reported in Section 3.
Table 2-1. Parameters for Major Contaminants and Spectroscopic Interferants
Chemical
Formula
H20
CO2
SF6
C3F7C(0)C2F5
S02
CO
HF
COF2
C2H2
C2H4
C2F6
CF4
CHF3
CH3F
CH4
CH20
CH202
NO
N20
N02
H2SO4
SO3
Compound
Water
Carbon Dioxide
Sulfur Hexafluoride
Novec™ 612
Sulfur Dioxide
Carbon Monoxide
Hydrofluoric Acid
Carbonyl Fluoride
Acetylene
Ethylene
Hexafluoroethane
Carbon Tetrafluoride
Trifluoromethane
Methyl Fluoride
Methane
Formaldehyde
Formic Acid
Nitric Oxide
Nitrous Oxide
Nitrogen Dioxide
Sulfuric Acid
Sulfur Trioxide
SF6
(ppmv-meters)
2.89-22.3*
70-2,110
58 - 92,701
n/a
518-10,415
26 - 20,358
1 - 2,000
50 - 5,000
111-5,550
86 - 2,576
448-1,119
5.6-1,120
112-560
177-1,182
87-21,119
92-1,838
76
53 - 2,043
102-1,019
34 - 1,544
n/a
n/a
Novec™ 612
(ppmv-meters)
2.89-22.3*
137-510*
56 - 280
99 - 991
n/a
784 - 20,358
1 - 2,000
50 - 5,000
111-5,550
86 - 2,576
448-1,119
5.6-1,120
112-560
177-1,182
87-21,119
92 - 1,838
76
53 - 2,043
102-1,019
34 - 1,543
n/a
n/a
S02
(ppmv-meters)
2.89-22.3*
70-2,110
56 - 280
n/a
518-35,770
26 - 3,863
1 - 2,000
n/a
n/a
86 - 2,576
n/a
n/a
n/a
n/a
87-21,119
92 - 1,838
76
53 - 2,043
102-1,019
34 - 1,543
164
1,400
*Expressed in percent-meters since high concentration references were required.
2-2
-------
Sensitivities associated with the optimized analysis methods are highly field dependent
because they are subject to overall moisture content (which affects the IR detector signal-to-
noise ratio) and to sample matrix content (which adds interferences to quantification regions).
Therefore, detection limits are discussed in Section 3 and Section 4 as they relate to the field
measurements of low concentration byproducts and worker exposure compounds.
2.1.2. The Extractive FTIR Systems
Three extractive FTIR systems were used in this study. MKS (On-Line) FTIR
spectrometers were used to analyze the casting hood environments, while a Thermo-Nicolet
spectrometer was used to monitor the worker exposure location. Stainless steel sample probes
(Vi-inch out diameter (OD)) were used to extract gas samples via venturi pumps connected to the
exhaust of the short path (5.11 m) sample cell for the hot zone and of the long path (20.1 m)
sample cell for the cold zone. A long perfluoroalkoxy (PFA) Teflon line (H-inch OD) acted as
the sample probe at the worker exposure location. Flows on the order of 5 1pm were maintained
through each extraction system. The sample flow temperatures were maintained at 150 °C at the
casting hood locations and at room temperature (-30 °C) at the worker exposure location. The
casting hood FTIR systems maintained elevated temperatures for their sample cells and
extraction lines to preclude condensation loss and acid mist formation. There were two sampling
port locations for both the hot zone and cold zone casting hood areas - an upstream location
closest to the casting pours and a downstream location farthest from the casting pours. The
stainless steel probes within each hood area were moved from upstream locations to downstream
locations for periods of continuous sampling at each location. The probes extended about 6
inches into the casting hood. Approximate dimensions and configurations are indicated in the
sampling schematic (Figure 2-1).
Inside each FTIR cell, a set of optically matched gold-plated mirrors reflected an IR
beam through the sample gas multiple times. As the beam passes through the sample, the
molecules in the sample absorb some of its energy. After exiting the cell, the IR beam was
directed to a liquid-nitrogen cooled-mercury/cadmium/telluride (MCT) detector, which is a
photoconductive device that produces an electrical voltage proportional to the amount of IR
energy that strikes it. The magnitude of IR absorbance at particular frequencies is a measure of
the concentrations of chemical compounds. The cell path length is the total distance traveled by
the IR beam inside the cell and is an important variable used to determine sample concentrations.
For this project, cell path lengths were fixed at 20.1 m for the casting hood cold zone FTIR, 5.1
m for the hot zone FTIR, and 10 m for the worker exposure monitoring system. Interferometer
resolutions were set to 0.5 cm"1 and signal averaging was performed over two-min periods.
2-3
-------
worker exposure
sampling location
Worker area (enclosed)
(not to scale)
50-ft Heated Teflon
Metal pump
line
Approximate casting hood sampling locations
(~3 feet apart)
Figure 2-1. Casting Hood and Sampling System Schematic
2-4
-------
2.2. Principles of RGA Monitoring
A mass spectrum is obtained by converting components of a sample into rapidly moving
gaseous ions and resolving them on the basis of their mass-to-charge ratios. The principles of
mass spectrometry are straightforward; a block diagram showing the major components of the
URS-built RGA is displayed in Figure 2-2.
Sample
Chamber
Aperture
Inlet System
Ion Source
+
Mass Analyzer
+
Detector
Signal Processor
Vacuum
System
Vacuum
System
Data Archival and
Control
\
Figure 2-2. RGA Component Block Diagram
Descriptions of these components are included in Section 2.2.2. As molecules from the
sample are ionized in the analyzer chamber, the detector registers a response for a given mass-to-
charge (m/e) ratio at an intensity proportional to the absolute molecule count. The following
section reports the desired m/e ratios for argon and krypton, which were monitored during
dilution measurements.
2.2.1. The RGA Spectrum Analysis Method
Since the RGA quadrupole mass analyzer breaks down molecules (or, in this case, the
natural atomic species argon or krypton) into fragments of varying m/e ratios, the specific m/e
2-5
-------
for each compound of interest that leads to the greatest response at the detector was targeted.
Table 2-2 lists the relative isotopic abundances in nature for argon and krypton.
Table 2-2. Relative Isotopic Abundances for Argon and Krypton
Isotope
36-Ar
38-Ar
40-Ar
78-Kr
80-Kr
82-Kr
83 -Kr
84-Kr
86-Kr
Accurate Mass
(m/e, unitless)
35.9675456
37.9627322
39.9623831
77.9203970
79.9163750
81.9134830
82.9141340
83.9115060
85.9106140
Abundance
(percent)
0.3380
0.0630
100.0000
0.6140
3.9474
20.3509
20.1754
100.0000
30.3509
According to the information in Table 2-2, the derived m/e value for the "parent" argon
ion is 40 and for krypton is 84. As a result, rather than scanning across each m/e channel within
its measurement range of 2 to 100 amu, the analyzer was focused directly to either m/e = 40 or
m/e = 84. A few other m/e values were scanned during monitoring periods for diagnostic
purposes, including parent ions for nitrogen, oxygen, water, and carbon dioxide. In order to
enable measurements of dilution percentage, the RGA detector response is calibrated
periodically by relating known argon and krypton concentrations to the detector signal. RGA
calibrations are reported in Appendix A and the dilution determinations are reported in Section 4.
2.2.2. The Extractive RGA System
Traditionally, mass spectrometers are used in a vacuum. Coupled with the development
of atmospheric samplers and closed ion sources, recent advancements of this technology have
enabled atmospheric sampling. The "high pressure" URS-built RGA is smaller, more robust,
and much more portable than its laboratory predecessors. Gas samples were brought to the mass
analyzer vacuum chamber as slip-streams from the FTIR sample inlets via a venturi pump. The
pump provided the sample matrix at less-than-atmospheric (-500 torr) pressures and with small
residence times within the sample line tubing. The venturi pump extraction region, which is
basically the inlet of the pump, was interfaced to the RGA sample chamber with a small orifice
valve that was manually tuned to maintain a reasonable leak rate into the sample chamber. This
orifice valve was separated from the mass analyzer chamber by a small diameter (< 100 |j,m)
aperture. The inlet system was set to maintain a sample pressure of ~5 x 10"4 torr, which was an
increase of two orders of magnitude over the mass analyzer chamber background pressure and
which was maintained by turbomolecular pump. Given an argon or krypton tracer concentration
2-6
-------
of approximately 1 percent (10,000 ppmv), a detection sensitivity of approximately 100 ppmv
was achieved. Such a detection sensitivity enabled the accurate measurement of dilution
percentage values up to 99 percent (1-[100 ppmv / 10,000 ppmv] * 100).
The RGA detection system housed in the main (mass analyzer) chamber was a
Micropole™ mass analyzer manufactured by Horiba. It consisted of an integrated package that
operated a tungsten filament (the ionizer) coupled to a series of focusing lenses and a miniature
array of quadrupoles (the mass analyzer) that allowed the ions to strike a Faraday Cup (the
detector). The mass range for this analyzer was 2 to 100 amu with a resolution of 1 amu. A
RS232 digital interface to a laptop and appropriate software allowed continuous operation and
data archival.
2.3. Ambient Air Dilution Considerations
Though the ingot casting machine hot and cold zones were somewhat contained in a
hooded enclosure, they were not completely sealed. A significant amount of ambient air dilution
was anticipated and must be considered when computing destruction rates based on
concentration measurements. As a result, the ambient air dilution within the casting hood was
experimentally determined using three distinct approaches.
1) Ambient argon intrusion: This was anticipated to provide the most direct and
continuous measurement approach because the concentration of argon was tracked
in real time via RGA at the same extractive sample locations as the FTIR systems.
Ambient levels of argon were assumed to be on the order of 1 percent (10,000
ppmv) and were factored into the dilution calculations when known flow rates of
compressed dry air (CDA) from the direct cover gas mixture were considered
supplemental to air dilution. Therefore, given the field RGA sensitivity for argon on
the order of 100 ppmv, dilution rates up to 99 percent could be tracked. With only
one RGA system available for sampling, the hot casting zone was monitored during
the first half of each cover gas testing period and the cold casting zone was
monitored during the second half of the period.
2) Krypton tracer: A krypton tracer study was intended to further support the direct
argon intrusion measurements. An added benefit of this part of the study was that
krypton background concentrations and cover gas mixture contributions were
negligible, which minimizes dilution rate bias. One point of concern in this analysis
was the possibility of krypton interacting with free fluorine in the thermal plasma at
the magnesium melt surface. The heavy noble gas xenon will readily form
complexes with fluorine and oxygen, while the reactivity of krypton is less known
but is expected to be weaker than with xenon.
2-7
-------
3) Cover gas measurements during non-casting periods: Because casting operations at
MagReTech were constant while the cover gas was being applied, measurements
taken from the casting hood during normal testing conditions (a moving ingot mold
conveyer belt with metal present) could be compared to a situation with a moving
belt but no metal present. Since no magnesium was present in the casting hood
zones to react with or degrade the cover gas, any reduction in the concentration of
the cover gas constituents would be solely attributable to ambient air dilution. This
test was run once for each cover gas system examined. A potential anticipated
drawback to this procedure was that the ambient air/cover gas dilution dynamics
may be different in a casting hood without molten magnesium and its resulting
convective effects, though ingot molds are pre-heated. The dilution estimates
determined through this approach were then used to determine cover gas
destruction.
Injecting argon tracer gas directly into the cover gas mixture manifold at concentrations
much greater than ambient levels would further support the results obtained from the ambient
argon intrusion monitoring. High tracer concentrations are needed to overcome not only the
native amounts of argon present due to significant ambient air dilution but also the argon already
present in the CDA. High concentrations of inert tracer gas would potentially have an adverse
impact on the cover gas mixture needed for processing ingots by displacing the active cover gas
constituents. Therefore, only the three approaches described above were carried out, and the
results of these approaches are reported in Section 4.
2-8
-------
3. Monitoring Results
3.1. Casting Hood Monitoring
Each cover gas mixture has the potential to generate a variety of chemical byproducts that
are due to local thermal plasma effects near the ingot melt surfaces in conjunction with
significant amounts of ambient moisture within the casting hood environment. The amount of
air dilution was expected to impact the type and relative amounts of these byproducts, but the
extent of this impact is not completely understood. For example, air dilution provides a source
of hydrogen as a chemical pathway so that all fluorinated cover gas mixtures (SF6/SO2 and
Novec™ 612) were expected to produce a hydrogen fluoride (HF) byproduct. Also, the thermal
plasmas within the local volumes around each ingot mold were expected to break down the base
cover gas mixture components into reactive atomic and free radical species. These species
would then recombine into other byproducts that were mostly fluorinated. This would be
especially true for Novec™ 612 mixtures, as perfluorocarbons (PFCs) have previously been
observed during tests conducted on magnesium die casting holding furnaces.5
However, for this project, the expected byproducts normally produced at low ppmv levels
were not analytically detected because of the small surface areas of the magnesium melts and the
high degree of dilution in the ingot casting hood. Table 3-1 summarizes the cover gas, ambient
air, and combustion-type compounds and the expected destruction byproducts for the SF6/SO2
cover gas runs. The minimum detection limits (MDLs) for the anticipated byproducts are also
reported. These MDLs were considered to be field-representative by taking three times the
standard deviation of the quantitative analysis values for each compound that were scattered
about zero. The sample spectra data sets were validated by confirming that they contained all the
primary infrared absorption interferences (for H^O, CC>2, base cover gas compounds, etc.) but
none of the byproducts listed as not detected (ND) in the tables below. Tables 3-2 and 3-3
summarize the observable compounds for the Novec™ 612 cover gas runs. In tests 5hd and 6hd,
CHF3 was measured at extremely low levels, averaging at or below the MDL. Table 3-4
summarizes the observables for the SC>2 cover gas mixture. Sulfuric acid (H^SC^) was
anticipated but not detected throughout both sampling periods.
Special attention was paid to perfluoroisobutylene (PFIB) as one potential byproduct of
the Novec™ 612 cover gas mixture. It was important to monitor PFIB from an occupational
exposure standpoint because it is has extremely low exposure limits. A suitable FTIR spectral
reference was unattainable, and the URS laboratory was not able to generate a reference because
PFIB is a controlled substance. However, a mass spectrum of this gas has been published and
was referenced on-site against a continuous two-hour block of full-spectral RGA scans taken
during the Novec™ 612 testing. The appearance of RGA peaks at m/e = 69, 31, and 93 at
5 US EPA. Characterization of Emissions and Occupational Exposure Associated with Five Cover Gas
Technologies for Magnesium Die Casting, 2007
5-1
-------
relative abundances of 100 percent, 60 percent, and 30 percent, respectively, would have been an
indicator of the presence PFIBs. However, PFIB was not detected during this study because
signals at m/e = 69 were never observed to be above noise levels. The estimated detection limit
for partial pressures at m/e = 69 was on the order of 10 ppmv.
Table 3-1. Data Summary for SF6 / SO2 Cover Gas Mixture
Date
9/10/07
9/11/07
9/13/07
Test
Location
Direct
Ihu
leu
Direct
2hu
2cu
Direct
3hu
3cu
3hd
3cd
Min
Max
Avg
MDL
Min
Max
Avg
MDL
Min
Max
Avg
MDL
Min
Max
Avg
MDL
Min
Max
Avg
MDL
Min
Max
Avg
MDL
Min
Max
Avg
MDL
Min
Max
Avg
MDL
SF6
(ppmv)
10,675
1,332
1,478
1,407
507
646
572
7,554
1,016
1,418
1,251
375
537
456
8,463
648
1,651
1,232
442
865
639
283
565
424
236
499
333
SO2
(ppmv)
4,039
502
547
530
210
265
233
8,249
605
1,127
724
199
411
232
6,682
449
1016
843
298
557
446
143
322
216
126
396
187
H2O
(%)
1.04
1.25
1.09
0.94
1.07
0.97
1.18
1.49
1.34
0.79
0.96
0.88
1.07
1.54
1.27
0.58
1.03
0.75
1.11
1.48
1.28
0.73
0.85
0.78
C02
(%)
6.21
6.56
6.42
2.36
2.87
2.67
6.18
7.52
7.12
1.85
2.57
2.11
3.86
8.32
6.81
1.72
3.08
2.64
242
3.59
2.91
1.53
2.58
1.92
CO
(ppmv)
0.41
1.95
0.83
0.50
2.14
0.96
0.20
5.88
0.56
0.17
4.04
0.48
0.43
64.6
10.1
0.34
138
32.9
0.27
1.64
0.62
0.33
2.21
0.72
CH4
(ppmv)
2.54
10.0
3.79
2.40
9.63
3.81
2.19
3.61
2.66
2.10
3.41
2.61
2.54
371
56.1
1.91
354
82.8
2.14
9.87
3.24
1.49
4.42
2.68
HF
(ppmv)
0.05
0.17
0.09
0.26
0.30
0.28
0.02
0.07
0.03
0.09
0.15
0.11
0.07
0.11
0.09
0.08
0.26
0.17
0.07
0.11
0.09
0.18
0.37
0.26
COS
(ppmv)
ND
ND
ND
0.01
ND
ND
ND
0.01
ND
ND
ND
0.03
ND
ND
ND
0.03
ND
ND
ND
0.03
ND
ND
ND
0.06
ND
ND
ND
0.01
ND
ND
ND
0.04
CS2
(ppmv)
ND
ND
ND
0.44
ND
ND
ND
1.17
ND
ND
ND
0.61
ND
ND
ND
2.11
ND
ND
ND
0.75
ND
ND
ND
0.76
ND
ND
ND
1.00
ND
ND
ND
0.34
H2S
(ppmv)
ND
ND
ND
8.46
ND
ND
ND
22.5
ND
ND
ND
15.5
ND
ND
ND
28.8
ND
ND
ND
19.2
ND
ND
ND
37.2
ND
ND
ND
18.6
ND
ND
ND
46.9
H2SO4
(ppmv)
ND
ND
ND
0.34
ND
ND
ND
0.34
ND
ND
ND
0.64
ND
ND
ND
0.49
ND
ND
ND
1.08
ND
ND
ND
1.06
ND
ND
ND
2.92
ND
ND
ND
1.30
MDL is reported if the compound was not detected (ND).
5-2
-------
Table 3-2. Data Summary for MTG-Shield™ using Novec™ 612
Date
9/11/07
Test
Location
Direct
4hu
4cu
Direct
5hd
5cd
Min
Max
Avg
MDL
Min
Max
Avg
MDL
Min
Max
Avg
MDL
Min
Max
Avg
MDL
Novec™
612
(ppmv)
2,000
196
247
219
70.4
86.4
76.9
1,500
43.7
49.4
47.0
26.3
27.9
27.0
H20
(%)
1.00
1.28
1.15
0.76
0.91
0.84
1.02
1.14
1.07
0.77
0.84
0.80
C02
(%)
6.82
8.78
7.81
2.21
2.86
2.59
2.24
2.50
2.40
1.30
1.34
1.31
CO
(ppmv)
1.09
2.44
1.94
0.87
1.73
1.40
0.81
1.12
0.95
0.59
0.84
0.72
CH4
(ppmv)
2.84
4.55
3.74
4.39
5.99
5.17
3.27
4.09
3.69
4.71
5.96
5.31
HF
(ppmv)
0.11
0.24
0.17
0.22
0.35
0.31
0.12
0.17
0.15
0.19
0.23
0.22
SF6
(ppmv)
0.49
0.77
0.64
0.65
0.96
0.79
0.57
0.73
0.64
0.79
0.91
0.86
CHF3
(ppmv)
ND
ND
ND
0.16
ND
ND
ND
0.27
ND
0.30
0.15
0.15
ND
ND
ND
0.37
CF4
(ppmv)
ND
ND
ND
1.34
ND
ND
ND
0.63
ND
ND
ND
2.05
ND
ND
ND
1.36
C2F6
(ppmv)
ND
ND
ND
0.67
ND
ND
ND
0.06
ND
ND
ND
0.17
ND
ND
ND
0.12
C3F8
(ppmv)
ND
ND
ND
0.24
ND
ND
ND
0.08
ND
ND
ND
0.25
ND
ND
ND
0.09
COF2
(ppmv)
ND
ND
ND
0.18
ND
ND
ND
0.16
ND
ND
ND
0.18
ND
ND
ND
0.33
CH202
(ppmv)
ND
ND
ND
0.69
ND
ND
ND
0.21
ND
ND
ND
0.45
ND
ND
ND
0.15
H2CO
(ppmv)
ND
ND
ND
0.05
ND
ND
ND
0.10
ND
ND
ND
0.06
ND
ND
ND
0.17
MDL is reported if the compound was not detected (ND).
3-3
-------
Table 3-3. Data Summary for MTG-Shield™ using Novec™ 612
Date
9/12/07
9/12/07
Test
Location
Direct
6hu
6cu
6hd
6cd
Direct
7hu
7cu
Min
Max
Avg
MDL
Min
Max
Avg
MDL
Min
Max
Avg
MDL
Min
Max
Avg
MDL
Min
Max
Avg
MDL
Min
Max
Avg
MDL
Novec™
612
(ppmv)
1,102
77.5
112
95.6
137
223
161
34.9
41.3
37.5
52.1
84.1
73.8
1,663
85.7
158
131
203
260
228
H20
(%)
0.96
1.28
1.12
2.46
2.97
2.67
1.15
1.58
1.30
2.60
3.63
2.86
0.98
1.21
1.10
2.60
3.31
2.91
C02
(%)
4.79
7.16
6.66
8.33
11.0
9.17
2.99
3.61
3.22
4.79
5.24
4.99
3.93
6.35
5.25
7.30
9.47
8.44
CO
(ppmv)
0.45
3.35
1.66
0.72
9.78
2.44
0.61
105
10.7
0.87
436.5
36.8
0.26
0.93
0.52
0.62
2.57
1.20
CH4
(ppmv)
1.77
2.86
2.12
3.56
10.7
7.45
2.09
719
71.1
10.2
1558
139
1.32
2.32
1.82
4.57
9.85
6.93
HF
(ppmv)
0.05
0.16
0.09
0.35
0.82
0.61
0.14
0.18
0.17
0.73
1.29
0.97
0.12
0.38
0.26
0.85
1.47
1.16
SF6
(ppmv)
0.07
0.22
0.11
0.28
1.09
0.50
0.08
0.32
0.16
0.63
1.75
0.99
0.04
0.13
0.08
0.09
0.52
0.28
CHF3
(ppmv)
ND
ND
ND
0.09
ND
ND
ND
0.90
ND
0.30
0.15
0.22
ND
ND
ND
1.92
ND
ND
ND
0.22
ND
ND
ND
0.60
CF4
(ppmv)
ND
ND
ND
1.58
ND
ND
ND
5.21
ND
ND
ND
1.48
ND
ND
ND
7.95
ND
ND
ND
1.48
ND
ND
ND
3.10
C2F6
(ppmv)
ND
ND
ND
0.41
ND
ND
ND
0.81
ND
ND
ND
1.10
ND
ND
ND
0.39
ND
ND
ND
1.10
ND
ND
ND
0.40
C3F8
(ppmv)
ND
ND
ND
0.29
ND
ND
ND
0.26
ND
ND
ND
0.64
ND
ND
ND
0.22
ND
ND
ND
0.64
ND
ND
ND
0.19
COF2
(ppmv)
ND
ND
ND
0.16
ND
ND
ND
0.53
ND
ND
ND
0.17
ND
ND
ND
0.53
ND
ND
ND
0.17
ND
ND
ND
0.28
CH202
(ppmv)
ND
ND
ND
0.76
ND
ND
ND
1.17
ND
ND
ND
1.12
ND
ND
ND
0.54
ND
ND
ND
1.12
ND
ND
ND
0.71
H2CO
(ppmv)
ND
ND
ND
0.05
ND
ND
ND
0.61
ND
ND
ND
0.05
ND
ND
ND
0.66
ND
ND
ND
0.05
ND
ND
ND
0.20
MDL is reported if the compound was not detected (ND).
3-4
-------
Table 3-4. Data Summary for SO2 Cover Gas Mixture
Date
9/13/07
9/14/07
Location
Direct
Shd
Scd
Direct
9hu
9cu
Min
Max
Avg
MDL
Min
Max
Avg
MDL
Min
Max
Avg
MDL
Min
Max
Avg
MDL
S02
(ppmv)
10,000
221
295
277
103
142
128
12,000
539
1,077
946
263
320
296
H2O
(%)
0.89
1.13
0.95
0.75
0.84
0.80
1.08
1.34
1.22
0.67
0.82
0.75
C02
(%)
2.20
2.40
2.27
0.94
1.13
0.99
0.04
7.21
1.06
0.04
0.04
0.04
CO
(ppmv)
0.14
0.66
0.28
0.16
0.43
0.29
0.23
1.95
0.53
0.28
4.30
0.88
CH4
(ppmv)
3.32
3.99
3.62
2.82
4.01
3.18
2.67
4.03
3.40
2.94
4.09
3.42
COS
(ppmv)
ND
ND
ND
0.004
ND
ND
ND
0.02
ND
ND
ND
0.01
ND
ND
ND
0.03
CS2
(ppmv)
ND
ND
ND
3.54
ND
ND
ND
2.34
ND
ND
ND
2.09
ND
ND
ND
1.20
H2S
(ppmv)
ND
ND
ND
24.6
ND
ND
ND
33.4
ND
ND
ND
18.5
ND
ND
ND
19.3
H2SO4
(ppmv)
ND
ND
ND
0.05
ND
ND
ND
0.02
ND
ND
ND
0.07
ND
ND
ND
0.03
MDL is reported if the compound was not detected (ND).
3-5
-------
3.2. Worker Exposure Monitoring
The cover gases evaluated in this study can produce byproducts that may be of concern
from an occupational exposure standpoint. Therefore, a third extractive FTIR system was used
to monitor the ambient air in the casting machine operator room (see Figure 2-1) for any
potential occupational exposure hazards associated with the usage of each cover gas. For
example, SC>2 and FTP have very low eight-hour time-weighted average exposure limits of 2 and
3 ppmv, respectively.6 The breathing zone located near the worker responsible for controlling
ingot pours and skimming ingot surfaces was continuously monitored during the testing. Table
3-5 summarizes the concentrations observed, as well as the pertinent MDLs for the compounds
not detected, for those species present in the casting hood at the highest concentrations - namely
the primary cover gas compounds and the most significant byproduct (HF). Only SF6 was
observed, and only during its usage as a cover gas. The spectra were surveyed for the
appearance of features attributable to compounds outside of those listed in Table 2-1 but none
were observed besides expected ambient air constituents.
Table 3-5. Worker Exposure Monitoring
Date
(m/dd/yy)
9/10/07
9/11/07
9/13/07
9/11/07
9/11/07
9/12/07
9/12/07
9/13/07
9/14/07
Approx.
Casting Time
(Local Time)
1730 - 1930
1115- 1345
0915 - 1200
1500 - 1709
1710-1730
0915-1140
1420 - 1620
1340 - 1540
0740 - 1000
Cover Gas Mixture
Components
SF6/SO2/CDA/CO2
SF6/S02/CDA/C02
SF6/S02/CDA/C02
Novec™612/CDA/CO2
Novec™612/CDA/CO2
Novec™ 612 / CDA / CO2
Novec™ 612 / CDA / CO2
SO2/CDA
SO2/CDA
Average SF6
(ppmv)
0.22
0.19
0.57
n/a
n/a
n/a
n/a
n/a
n/a
Average SO2
(ppmv)
<0.5
<0.5
<0.5
n/a
n/a
n/a
n/a
<0.5
<0.5
Average
Novec™ 612
(ppmv)
n/a
n/a
n/a
<2.0
<2.0
<2.0
<2.0
n/a
n/a
Average
HF
(ppmv)
<0.1
<0.1
<0.1
<0.1
<0.1
<0.1
<0.1
<0.1
<0.1
Compounds listed with values as < X were not observed; their detection limits are reported as the value X.
n/a - not applicable
' OSHA Permissible Exposure Limits (PELs), http://www.osha.gov
J-C
-------
4. Cover Gas Destruction
Throughout each casting run listed in Table 1-1, the primary cover gas components and
byproducts were quantified simultaneously at both the casting hood hot zone and the cold zone
(see Figure 2-1). Roughly half of each monitoring period was spent sampling the upstream
ports; the other half was spent at the downstream ports. On some occasions, it was possible to
monitor at these sampling ports while the ingot casting conveyor was moving and still heated but
with no magnesium being poured: this was known as a "noncasting condition". In addition, on
some occasions it was possible to sample the cover gas composition at the outlet of the gas
blending manifold before injection into the casting hood. This was known as a "direct" cover
gas mixture measurement. Average concentrations over the sampling periods were then used to
calculate cover gas destruction percentages via the following approach:
Consider the injection cover gas concentration (after factoring in ambient air dilution)
versus the cover gas concentrations at the sampled locations. Calculate the destruction
factor (DF) as a percentage using
DF=WOx
sample cover gas cone, (ppm)
1 X
direct cover gas cone, (ppm) (1 _ DP
100.
where DP is the dilution percentage, which is determined experimentally by krypton
tracer or argon intrusion and casting vs. noncasting monitoring, as reported in Section
4.1.
The concentration and DF results for each cover gas mixture are reported in section 4.2.
4.1. Determining Dilution
Figures 4-1, Figure 4-2, Figure 4-3 plot the RGA-measured concentrations for argon and
krypton as the detector responses at m/e = 40 and m/e = 84 were isolated for each mass spectrum
and converted to concentrations via the appropriate calibration curve. Each graph corresponds to
a period of time when a krypton tracer was injected into the cover gas blending manifold with a
flow producing concentrations on the order of 0.5 - 1 percent. Therefore, dilution percentages
were determined redundantly by simultaneous measurement of both direct argon intrusion and
krypton tracer. Notated on each graph is when the monitoring occurred at a specific sampling
location (upstream hot zone, downstream hot zone, upstream cold zone, downstream cold zone,
and cover gas direct).
4-1
-------
RGA Dilution Determination: Kr Tracer and Direct Ar Monitoring
[SF6 / SO2 mix as cover gas; 9/11/07]
Sampling of direct
upstream hot zone location:
Dilution by direct Ar = 83%
Dilution by Kr tracer = 86%
upstream cold zone location
Dilution by direct Ar = 95%
Dilution by Kr tracer = 98%
0.0
0 1000 2000 3000 4000 5000 6000 7000 8000 9000 10000
Time elapsed (sec)
Figure 4-1. RGA Dilution Measurements, 11 September 2007
4-2
-------
o
c
o
O
RGA Dilution Determination: Kr Tracer and Direct Ar Monitoring
[Novec mix as cover gas; 9/12/07]
0.6-
< 0.4-
0.2-
• % Ar
• % Kr
Sampling of direct
cover gas mix
upstream hot zone location:
Dilution by direct Ar = 87%
Dilution by Kr tracer = 91%
upstream hot zone location:
Dilution by direct Ar = 94%
Dilution by Kr tracer = 91%
downstream hot zone location:
Dilution by direct Ar = 90%
Dilution by Kr tracer = 96%
upstream cold zone location:
Dilution by direct Ar = 98%
Dilution by Kr tracer = 97%
2000
4000 6000
Time elapsed (sec)
8000
10000
Figure 4-2. RGA Dilution Measurements, 12 September 2007
4-3
-------
RGA Dilution Determination: Kr Tracer and Direct Ar Monitoring
[SF6 / SO2 mix as cover gas; 9/13/07]
1.2-
0.2-
0.0-
• % Ar
• % Kr
upstream hot zone location:
Dilution by direct Ar = 95%
Dilution by Kr tracer = 91%
cold zone locatior
\. Dilution by Kr trad
\
downstream hot zone locatio
Dilution by direct Ar = 99%
Dilution by Kr tracer = 96%
2000
4000 6000 8000
Time elapsed (sec)
Sampling of direct
cover gas mix
10000
Figure 4-3. RGA Dilution Measurements, 13 September 2007
DP calculations were carried out accordingly.
For krypton tracer,
= \OOx
I —
sample krypton (%
direct krypton in cover gas mixture (%
• For argon intrusion, the calculations were more complex because the levels of argon
native in the ambient air7 and already present in the cover gas mixture from blended
CDA must be considered. The equation below should be solved for DP.
DP
sample arg on (%) = x (arg on in ambient air, 0.9332% )
1 | x (arg on direct from cov er gas (%))
7 S. Y. Park, et al., Metrologia 41,387-395 (2004)
4-4
-------
The DP values at each location are reported in Table 4-1 along with their measurement
uncertainties. Uncertainties were derived based on the measurement standard deviations across
each monitoring period, which was assumed to take into account intermediate sampling errors
and analytical measurement variability since the casting process was run in a steady state. There
may also be inherent sampling variability due to ingot mold movement through the casting hood
and interactions with flows from cover gas nozzles. The standard deviations were propagated
throughout the DP calculation to produce the absolute measurement uncertainties listed in Table
4-1. The differences between DP values determined by krypton tracer and those determined by
argon intrusion were similar to measurement uncertainties.
Table 4-1. Dilution Percentages (DP) Calculated by Kr Tracer and Direct Ar Monitoring
Calculation Method
Kr Tracer
Direct Ar
Date
9/11/2007
9/12/2007
9/13/2007
9/11/2007
9/12/2007
9/13/2007
huDP
(percent)
86%(13)
91%(16); 91%(17)
91%(17)
83%(16)
87%(21); 94%(23)
95%(20)
cuDP
(percent)
98%(20)
97%(21)
96%(21)
95%(18)
98%(25)
97%(20)
hdDP
(percent)
96%(24)
96%(23)
90%(23)
99%(22)
cdDP
(percent)
97%(25)
98%(23)
Parenthetical values represent (±) one absolute standard deviation.
Dilution percentages were also calculated using the concentrations measured directly at
the cover gas mixing point and the concentrations measured during noncasting conditions. DP
values calculated by this method are presented in Table 4-2 and used the following equation:
DP = 100-
noncasting feed gas cone, (ppmv)
direct feed gas cone, (ppmv)
Dilution estimates obtained using the direct and noncasting approach were used as the
primary factor for estimating dilution. For tests where this approach was not available due to a
lack of data, an average DP was created for each sampling point using the noncasting and Kr
tracer results. It should be noted that the results across the different dilution estimation
approaches are relatively consistent.
4-5
-------
Table 4-2. Dilution Percentages (DP) Calculated from Noncasting and Direct Concentrations
Test
1
2
3
4
5
6
7
8
9
hu
(percent)
86%(1)
83%(1)
-
89%(1)
-
-
92%(1)
-
-
cu
(percent)
94%(1)
94%(1)
-
96%(1)
-
-
86%(1)
-
-
hd
(percent)
-
-
-
-
97%(1)
-
-
-
-
cd
(percent)
-
-
-
-
98%(1)
-
-
-
-
"-" Indicates that either noncasting or direct concentration values are missing for that specific test
and sample point. Parenthetical values represent (±) one absolute standard deviation.
4.2. Determining Cover Gas Destruction
Table 4-3 presents the cover gas flow rate, delivery concentration, FTIR measured
concentration, dilution factor, and calculated DF value for each available sampling site for each
cover gas test. A noncasting run was not possible for one of the Novec™ 612 mixtures runs on
12 September 2007. For SC>2, a noncasting run was not conducted because of logistical reasons.
A direct sample of the 862 cover gas composition was also not possible from the temporary
setup used during processing, so the direct SC>2 concentrations were estimated by calculation
from the mass flow controller settings on the gas blending system. Average DF values for the
SFe, Novec™ 612, and SC>2 cover gas mixtures were 5 percent, 9 percent, and 39 percent,
respectively.
Determining DF values involved several experimental measurements to derive
concentrations and DP values. Each experimental measurement was subject to indeterminate
uncertainty and contributed to the indeterminate error of the final results. The errors were
propagated from measurements to final results by common rules that were derived from the total
differentiation (by sum of all partials) of the DF equations discussed at the beginning of Section
4.8 As will be discussed in Section 5, the DF values determined by dilution considerations carry
significant uncertainties because the DP values are rather large and contain significant variance.
This variance resulted in the generation of negative destruction values in some cases because the
destruction is very low and approaching zero. For the sake of clarity, the negative destruction
values are withheld from this report. The average destruction values presented are therefore only
based on the positive results calculated using this methodology.
' D. Skoog. Principles of Instrumental Analysis. 3rd Ed., CBS College Publishing, 1985
4-6
-------
Table 4-3. Percent Destruction for Cover Gas Testing
Test#
(site)
Ihu
leu
2hu
2cu
Shu
3cu
3hd
3cd
4hu
4cu
5hd
5cd
6hu
6cu
6hd
6cd
7hu
7cu
8hd
8cd
9hu
9cu
Cover Gas Mixture
Components
SF6/SO2/CDA/CO2
SF6/SO2/CDA/CO2
SF6/SO2/CDA/CO2
SF6/SO2/CDA/CO2
SF6/SO2/CDA/CO2
SF6/SO2/CDA/CO2
SF6/SO2/CDA/CO2
SF6/SO2/CDA/CO2
Novec™612/CDA/CO2
Novec™612/CDA/CO2
Novec™612/CDA/CO2
Novec™ 612/CDA/CO2
Novec™ 612/CDA/CO2
Novec™ 612/CDA/CO2
Novec™612/CDA/CO2
Novec™612/CDA/CO2
Novec™612/CDA/CO2
Novec™612/CDA/CO2
SO2/CDA
SO2/CDA
SO2/CDA
SO2/CDA
Flow3
(1pm)
270
135
198
99
270
135
270
135
189
95
189
95
191
96
191
96
189
95
191
95
191
96
Cover Gas
Delivery
Cone.
(ppmv)
10,675
10,675
7,554
7,554
8,463
8,463
8,463
8,463
2,000
2,000
1,500
1,500
1,102
1,102
1,102
1,102
1,663
1,663
10,000
10,000
12,000
12,000
Cover
Gas
Measured
Cone.
(ppmv)
1,407
572
1,251
456
1,232
640
424
333
219
77
47
27
96
161
37
74
131
228
277
128
946
296
Dilution
Percentage0
(percent)
86%
94%
83%
94%
89%
94%
96%
98%
89%
96%
97%
98%
89%
94%
96%
98%
92%
86%
96%
98%
89%
94%
Estimated
Cover Gas
Destruction
Factor
(percent)
4%
8%
3%
7%
=0%
=0%
=0%
=0%
=0%
=0%
=0%
10%
23%
=0%
4%
=0%
=0%
5%
22%
49%
30%
56%
a Approximate, estimated by reading flow rates on gas delivery manifold rotameters (uncalibrated). 2/3 of flow
went to the hot zone and 1/3 to the cold zone.
0 The dilution factor presented here comes from the noncasting/direct DP calculation for each specific test.
When this value was unavailable, the averaged value of all DP values (except those determined by the Direct Ar
method) for the designated sampling point was used and is presented in italics.
4-7
-------
5. Discussion
5.1. Cover Gas Test Observations
Compared to the die casting crucibles previously studied, the design of typical
magnesium ingot casting hoods suggested that the cover gas destruction would be low and
difficult to estimate due to increased ambient air dilution and variability. This assumption was
made due to the four leading factors listed below.
1. The overall surface area of molten magnesium to be covered in an ingot caster is
smaller than typical die casting crucibles. Assuming that about six ingot molds
containing molten metal are within the hot zone at the same time, and given that each
mold has a surface area of about 800 cm2, the total surface area of the covered molds
is about 4,800 cm2. A 1.2 meter diameter die casting crucible has a surface area of
about 12,000 cm2. A smaller covered surface area means that less of the cover gas
concentrations are being consumed on a percentage basis than with a larger covered
surface area.
2. The freshly poured ingots within the hot zone begin cooling immediately after
injection of molten magnesium. Alloying crucibles and holding furnaces must keep
the metal in a liquid state throughout processing. This implies that less cover gas will
interact and break down at the ingot surface as it cools, leading to lower destruction
rates.
3. The casting hood volumes are greatly affected by ambient air dilution. Excessive
dilution makes differential measurements difficult to carry out because the amount of
dilution must be precisely and consistently characterized.
4. The casting hood volumes are quite large and the ingot protection atmosphere is very
turbulent due to high cover gas flow rates and movement of the ingot molds. These
characteristics create difficulties for continuous real-time extraction of representative
gas samples, which makes analytical measurement precision challenging and greatly
influences destruction calculations.
These factors were born-out in the actual measurement results. Several reasonable
estimations and observable trends can be gleaned from the results reported in Section 4. These
observations would include the following:
• Destruction was generally low under all cases, with the exception of 862, which
exhibited the highest destruction percentages (39 percent).
• The casting-versus-noncasting destruction determinations carry significantly lower
measurement uncertainties and better reproducibility at each sampling location than do
the destruction values determined by dilution, where some test cases yielded unrealistic
negative values.
5-1
-------
• As expected, the most consistently reliable destruction percentages for both measurement
methods were calculated at the upstream hot zone sampling location.
For measurements during ingot casting, this study yields the following primary
recommendations for future research: (1) maximize the representativeness of concentration
analysis by setting up as many sampling points as possible for simultaneous gas extraction from
the casting hood, and then (2) characterize ambient air dilution effects by basing the destruction
calculations upon casting-versus-noncasting conditions. The extent to which (1) can be applied
is highly dependent upon logistical and process concerns. For this study it was not feasible to set
up more than two sampling ports per casting hood zone without interfering with process
activities such as metal pouring, ingot skimming, and conveyor belt operation. An additional
recommendation is to account for air turbulence effects during monitoring periods. For example,
extend casting and noncasting events over longer continuous blocks of time to help smooth out
the averaging and subsequent comparison of concentrations during the casting-versus-noncasting
conditions.
One benefit of the low destruction values and excessive air dilution is that the
concentrations of cover gas byproducts were often negligible within the casting hood and, by
extension, also within the operator room environment. The tables in Section 3 indicate that the
only measurable byproduct was HF and its average concentrations were almost always under 1
ppmv. As expected, Novec™ 612 usually produced slightly more FTP than SF6/SO2, but no other
fluorinated species (including PFIB) were detected.
This study also indicates that there is significant uncertainty regarding the exact mixture
of cover gas being applied using the current rotameter-based control system. Measured cover
gas mixture concentrations for SF6 and SO2 from direct FTIR sampling at the manifold were
significantly higher than what was expected based on rotameter readings that controlled the
mixture. Monitoring of the cover gas system currently utilized at the facility indicates that there
may be significant over-protection occurring and that optimization to minimize cover gas usage
would be achievable if the current rotameter-based control system was replaced.
It should also be noted that the SF6/SO2 cover gas system is unique in that there are two
reactive cover gas constituents present. The exact nature of how this impacts the destruction of
SF6 is unclear. It is possible that a SF6-only cover gas system used in this application would
produce different results for destruction and byproduct formation.
5-2
-------
5.2. Climate Change Potential Discussion
One of the benefits of using Novec™ 612 and SC>2 as cover gases for magnesium melt
protection is their contribution to global climate change is significantly lower when compared to
SFe. This is evident when comparing their estimated global warming potentials (GWPs). Table
5-1 presents GWPs of several compounds detected during this study.
Table 5-1. Comparison of 100-Year GWP Estimates from the
Intergovernmental Panel on Climate Change (IPCC)
Fourth (2007) Assessment Report
Gas
Methane
Nitrous Oxide
Sulfur Hexafluoride (SF6)
IPCC GWP
25
298
22,800
IPCC (2007), Climate Change 2007: The Scientific of Climate Change.
Intergovernmental Panel on Climate Change, Cambridge University Press.
Cambridge, U.K.
The ingot casting area contains a variety of compounds, but only those with
corresponding GWP values were used in estimating the overall GWP impact of switching to
alternate cover gases. This calculation consisted of multiplying the average concentrations (parts
per million by volume) for each of the component cover gases and applicable destruction
products, by their respective GWP factors (obtained from the Second Assessment Report of the
Intergovernmental Panel on Climate Change) to obtain a GWP-weighted value. The summation
of all the GWP-weighted values for a particular cover gas resulted in the normalized CC>2
equivalent, which was compared to the CC>2 equivalent corresponding to SFe.9
Table 5-2 shows that when comparing the normalized CC>2 equivalent, or composite
GWP, the alternate cover gases have a much lower impact. An obvious source for this reduction
can be found in the incredibly high GWP of SFe shown in Table 5-1. Novec™ 612's GWP has
not been supplied by the IPCC, but is likely to be extremely low (i.e., Novec™ 612 is a
fluorinated ketone, which is assumed to have an atmospheric lifetime of approximately 5 days
and a GWP on the order of 1).10 862 is not an IR absorber and therefore has no global warming
potential. Compared to using SFe, switching to Novec™ 612 produces a reduction in overall
global warming impact of at least 99.7 percent.u Changing the cover gas from SF6 to 862
reduces the global warming impact by at least 99.9 percent but introduces a more complex
operational scenario due to toxicity concerns.
The above comparison does not include the specific flow rates for each cover gas. In
order to provide a more comprehensive analysis of composite GWP, an additional comparison
was conducted. Using the ideal gas law, the molecular weights of each gas and the delivery flow
9 Fugitive SF6 and CH4 concentrations were excluded from the overall GWP calculations for the cover gases.
10 D'Anna B, Sellevag S.R., Wirtz K., and Nielsen C.J. Photolysis Study of Perfluoro-2-methyl-3-pentanone Under
Natural Sunlight Conditions Environ Sci and Tech 2005 39(22) 8708-8711
5-3
-------
rate of the cover gas was used to estimate the composite emission rate in grams per hour (g/hr).
This equation can be described as follows:
Emission Rate — = ppmv x MW x Ipm x ^-(38.6 liters I mole x 106 J
^ hour ) hour
ppm = measured average concentration in parts per million
MW = molecular weight in grams per mole
1pm = gas flow in liters per minute
These values were summed to provide a composite GWP value that was weighted by the
cover gas flow rate. The average flow weighted GWP values were then compared against the
corresponding values for the SFe/SO2 system. Based on this approach, both of the cover gas
alternatives were observed to reduce GHG emissions by at least 99.9 percent relative to SF6/SO2
system. This result is also bourn out when comparing individual tests, such as Test 3 and 6;
averaging the results for the four monitoring points in each test results in a reduction in GHG
emissions of more than 99.9 percent. Details of the flow-weighted GHG emission impacts are
presented in Table 5-3.
5.3. Uncertainty Discussion
The results of this measurement study should not be interpreted to represent an absolute
analysis of GHG emissions associated with Novec™ 612, 862, and SF6 cover gas usage. While
this study does present a relatively accurate measurement analysis and approximate comparison
of GHG emissions, there are several areas of uncertainty inherent with this methodology. These
areas of uncertainty include FTIR and RGA error, error associated with blending gases, dilution
correction, and analytical and operational variation of the ingot casting machine evaluated. The
high levels of dilution - on the order of 90 percent - results in significant uncertainty associated
with destruction estimates.
Measurements taken by the FTIR and RGA are subject to variability inherent with highly
complex analytical equipment. While all prudent steps were taken during the measurement study
to minimize this contributor to uncertainty (see Section 2 and Appendix A), a small degree of
error is unavoidable.
11 Please refer to Section 5-3 for a discussion regarding the uncertainty associated with this methodology.
5-4
-------
Table 5-2. Normalized GWP Comparison of Measured Emissions from Inside the Casting Hood
Test#
(site)
Ihu
leu
2hu
2cu
Shu
3cu
3hd
3cd
4hu
4cu
5hd
5cd
6hu
6cu
6hd
6cd
7hu
7cu
8hd
8cd
9hu
9cu
Cover Gas Mixture
Components
SF6/S02/CDA/C02
SF6/S02/CDA/C02
SF6/SO2/CDA/CO2
SF6/S02/CDA/C02
SF6/S02/CDA/C02
SF6/SO2/CDA/CO2
SF6/SO2/CDA/CO2
SF6/S02/CDA/C02
Novec™ 612/CDA/CO2
Novec™ 612/CDA/CO2
Novec™ 612/CDA/C02
Novec™ 612/CDA/CO2
Novec™ 612/CDA/CO2
Novec™ 612/CDA/C02
Novec™ 612/CDA/C02
Novec™ 612/CDA/CO2
Novec™ 612/CDA/C02
Novec™ 612/CDA/C02
SO2/CDA
SO2/CDA
S02/CDA
S02/CDA
Cover Gas
Delivery Conc.a
(ppmv)
10,675
10,675
7,554
7,554
8,463
8,463
8,463
8,463
2,000
2,000
1,500
1,500
1,102
1,102
1,102
1,102
1,663
1,663
10,000
10,000
12,000
12,000
Measured Cover
Gas Cone.
(ppmv)
1,407
572
1251
456
1232
640
424
333
219
77
47
27
96
161
37
74
131
228
Til
128
946
296
GWP
Weighted
Cover Gas"
32,083,235
13,042,073
28,515,105
10,405,526
28,089,380
14,581,075
9,673,990
7,594,120
219.3
76.9
47.0
27.0
95.6
161.4
37.5
73.8
131.4
228.5
0
0
0
0
GWP
Weighted
CO2
64,180
26,674
71,212
21,135
68,095
26,374
29,109
19,233
78,066
25,915
23,956
13,137
66,593
91,743
32,158
49,905
52,494
84,434
22,754
9,905
10,585
434
GWP
Weighted
CH4
95
95
66
65
1,403
2,069
81
67
93
129
92
133
53
186
1,777
3,480
45
173
91
80
85
86
GWP
Weighted
SF6
0
0
0
0
0
0
0
0
14,666
18,084
14,518
19,682
2,450
11,385
3,639
22,514
1,794
6,474
8,545
11,086
6,981
8,061
Normalized CO2
Equivalent °
32,147,415
13,068,748
28,586,318
10,426,661
28,157,476
14,607,449
9,703,100
7,613,353
78,285
25,992
24,003
13,164
66,689
91,905
32,195
49,979
52,626
84,663
22,754
9,905
10,585
434
Average by
Cover Gas
18,038,815d
51,950
10,919
Chg from
SF6
(percent)
99.7%
99.9%
a Measured directly at cover gas manifold
bGWP weighting based on dilution corrected concentration for the primary cover gas constituent (e.g., Novec™ 612, SF6)
0 Please note that the normalized equivalent values exclude fugitive CH4 and SF6 emissions which are reported in italics for completeness.
d SF,; composite GWP baseline estimate for comparison with other tests.
5-5
-------
Table 5-3. GWP (Weighted by Cover Gas Flow) Comparison of Measured Emissions from Inside the Casting Hood
Test#
(site)
Ihu
leu
2hu
2cu
Shu
3cu
3hd
3cd
4hu
4cu
5hd
5cd
6hu
6cu
6hd
6cd
7hu
7cu
8hd
8cd
9hu
9cu
Cover Gas Mixture
Components
SF6/SO2/CDA/CO2
SF6/SO2/CDA/CO2
SF6/S02/CDA/C02
SF6/SO2/CDA/CO2
SF6/SO2/CDA/CO2
SF6/S02/CDA/C02
SF6/S02/CDA/C02
SF6/SO2/CDA/CO2
Novec™ 612/CDA/C02
Novec™ 612/CDA/C02
Novec™ 612/CDA/CO2
Novec™ 612/CDA/C02
Novec™ 612/CDA/C02
Novec™ 612/CDA/CO2
Novec™ 612/CDA/CO2
Novec™ 612/CDA/C02
Novec™ 612/CDA/CO2
Novec™ 612/CDA/CO2
S02/CDA
S02/CDA
SO2/CDA
SO2/CDA
Cover Gas Delivery
Conc.a
(ppmv)
10,675
10,675
7,554
7,554
8,463
8,463
8,463
8,463
2,000
2,000
1,500
1,500
1,102
1,102
1,102
1,102
1,663
1,663
10,000
10,000
12,000
12,000
GWP Weighted
Cover Gas"
(g/hr)
1,966,281
399,654
1,280,759
233,683
1,722,714
447,127
593,303
232,873
20
4
4
1
9
8
4
3
12
11
0
0
0
0
GWP
Weighted CO2
(g/hr)
1,185
246
964
143
1,259
244
538
178
1,010
168
310
85
871
600
421
326
679
546
297
65
138
3
GWP
Weighted CH,
(g/hr)
1
0
0
0
9
7
1
0
0
0
0
0
0
0
8
8
0
0
0
0
0
0
GWP
Weighted SF6
(g/hr)
0
0
0
0
0
0
0
0
629
388
623
422
106
247
158
488
77
139
370
240
303
175
Normalized CO2
GWP Equivalent0
(g/hr)
1,967,466
399,900
1,281,723
233,826
1,723,972
447,370
593,841
233,050
1,030
171
314
86
880
607
424
330
691
557
297
65
138
3
Average by
Cover Gas
(g/hr)
860,144d
480
126
Chg from SF6
(percent)
99.9%
99.9%
a Measured directly at cover gas manifold
b GWP weighting based on dilution corrected concentration for the primary cover gas constituent (e.g., Novec™ 612, SF6)
0 Please note that the normalized equivalent values exclude fugitive CH4 and SF6 emissions which are reported in italics for completeness.
d SF6 composite GWP baseline estimate for comparison with other tests.
5-6
-------
Appendix A - Calibrations and Diagnostic Checks
This section summarizes the on-site FTIR/RGA calibration and diagnostic procedures
carried out before and during the sampling tests.
A.1. FTIR Calibrations and System Checks
A series of on-site calibration and system checks was performed on each FTIR and
respective sampling system prior to testing to ensure data quality. These checks are described in
the remainder of this Section.
A.1.1. FTIR Sample Cell Integrity Checks
The integrity of each FTIR sample cell was confirmed prior to sampling by (1) drawing a
terminal vacuum of < 200 torr, then (2) sealing off the sample cell while still under vacuum, then
(3) monitoring any pressure rise (i.e., leak rate) within the cell by observing its pressure
transducer reading over a several-minute period. A cell was considered leak-tight when a leak
rate of < 2 torr min"1 was observed. The evacuated pressure on each FTIR sample cell (hot zone
system, cold zone system, and worker exposure system) did not rise above measurable values
over a 1-min period.
A.1.2. Infrared Detector Linearity Checks
For best results, the IR detector in each FTIR system must yield a linear response
throughout the measurement absorbance ranges within the measurement frequency range of all
sample spectra. A software linearizer was used to continuously adjust the MCT detector preamp
signal to achieve the desired linear response. To optimize the linearizer, background spectra
were acquired with and without a polystyrene film in the IR beam. Comparison of the strongly
absorbing polystyrene bands in the low-, mid-, and high-frequency regions against a clean
background enables the processor to appropriately set the linearizer terms (offset, linear, quad,
cubic, and delay). This procedure was run prior to the start of testing for each FTIR, and
subsequent spectra were periodically visually checked to confirm that linearity was maintained.
A.1.3. Noise Equivalent Absorbance (NEA, or Signal-to-Noise Ratio) Tests
NEA tests provide a measure of system noise - more specifically, the sensitivity of the
instrument at the specified spectral resolution (in this case, 0.5 cm"1) and number of co-added
spectra (in this case, 256, or 2 min of signal averaging). A two-min FTIR spectral background
was recorded while the sample cell was purged with dry nitrogen. A subsequent "sample"
spectrum was recorded while the cell was still under nitrogen purge immediately after the
background recording. The two spectra were ratioed to provide a snapshot of instrumental noise.
The NBAs of all three FTIR systems were well below 0.001 absorbance units across all
A-l
-------
measurement frequencies prior to sampling, which enabled instrument-limited quantitative
analysis sensitivities of < 1 ppmv to be achieved for all compounds of interest.
A.1.4. Path Length
The sample cells utilized for this study were geometrically fixed with an FTIR cold zone
system path length of 20.1 m and a hot zone system path length of 5.11 m. The worker exposure
FTIR system contained an adjustable multi-pass White cell that was aligned, set, and calibrated
at a path length of 10 m.
A.1.5. Spectrometer Frequency and Resolution Checks
A real-time check of frequency position and resolution was performed at each FTIR prior
to and directly following each round of testing. These checks were performed by monitoring a
specific water absorption band present in ambient air. The position of this line must not deviate
more than + 0.005 cm"1 from the reference value over the course of each test. Likewise, the
linewidth of this band, which is directly related to instrument resolution, must not deviate more
than + 0.05 cm"1 from the reference value over the course of each test.
A.1.6. Spectral Background
A spectral background is essentially a "blank spectrum" in that it does not contain any of
the target compounds normally present in the sample. It was created by purging each cell with
ultra-high-purity (UUP) nitrogen while recording a spectrum. This spectrum was then used by
the analytical software to ratio against each sample spectrum to produce an absorbance spectrum
for quantitative analysis. A new spectral background was generated each day prior to testing.
A.1.7. Sample Cell Exchange Rate
With sampling flow rates on the order of 5 L min"1 through each cell, complete sample
exchanges took approximately 7 s for the 5.11 m cell, 18s for the 20.1 m cell, and 30 s for the 10
m worker exposure cell, which had the largest internal volume. Since spectral signal averaging
was performed over 2-min intervals, each recorded spectrum represented an integrated average
over multiple sample cell exchanges.
A.2. RGA Calibrations and System Checks
A series of on-site calibration and system checks was performed on the RGA and
sampling system prior to and during sampling to ensure high data quality. These checks and
calibrations are described in the remainder of this Section.
A.2.1. Sample Inlet and Mass Analyzer Chamber Pressures
Pressure was continuously monitored in the mtorr range via a thermocouple gauge within
the sample inlet chamber, which is considered to be the high pressure side of the chamber
A-2
-------
aperture (see block diagram, Figure 2-2). The sample inlet chamber was directly interfaced to
the venturi pump-driven sample extraction line. The venturi pumping speed and valve orifice
maintained a constant pressure of 5 x 10"2 torr within the sample inlet chamber. Given a fixed
chamber aperture previously installed at URS, the 5 x 10"2 torr constant sample inlet chamber
pressure created a 5 x 10"4 torr total pressure within the mass analyzer chamber. Pressures were
continuously monitored by cold cathode gauge. When isolated from the sample inlet, total
background chamber pressures (~5 x 10"6 torr) were two orders of magnitude less than this mass
analyzer chamber total pressure. These pressures allowed RGA sensitivities for argon of-100
ppmv at m/e = 40, as previously mentioned in Section 2.2.2. The partial background pressures
observed for krypton at m/e = 84 were about half an order of magnitude lower than those
observed for argon at m/e = 40 (~5 x 10"7 torr versus ~1 x 10"6 torr), so the detection limit for
krypton was expected to be less than the 100-ppmv sensitivity noted for argon.
A.2.2. RGA Response Calibrations for Argon and Krypton
RGA responses at the mass analyzer channels corresponding to m/e = 40 and m/e = 84
were calibrated on two separate occasions (11 September 2007 and 13 September 2007) against
known concentrations of argon and krypton, respectively. Pure argon and krypton gases were
mixed with UHP nitrogen at precision flows delivered by mass flow controllers. The mixture
was sent to the sample inlet venturi pump under the same flow and orifice settings as when field
samples were collected. The calibration curves displayed in Figures A-l and A-2 were fit to the
detector signal and expressed as the ratio of partial pressure (percentage of total chamber
pressure at m/e = 40 or 84) to concentration (percentage). The curves were mainly linear at
lower concentrations with some nonlinearity at higher concentrations; therefore, they were fit to
second order polynomials with relatively small quadratic coefficients. The polynomials did not
differ a great deal between the first and second calibrations for each compound, so the calibration
curves produced on 11 September 2007 were applied to sample scans collected on 11-12
September 2007, whereas the calibration curves produced on 13 September 2007 were applied to
sample scans collected on 13-14 September 2007.
A-3
-------
(a)
9/11/07 Argon Calibration
1.0-,
0.8-
O 0.6'
c
o
O
< 0.4.
T3
CD
2
.Q
^ °'2'
O
0.0-
Y =-0.03201+0.27458 X+0.07741 X
0.0 0.5 1.0 1.5 2.0
RGA Detector Response (Pm/e=40 / P/100)
(b) 9/13/07 Argon Calibration
2.5
1.0-1
0.8-
^ 0.6-
O
O
o
< 0.4-
T3
CD
^ 0.2-
ca
O
0.0-
Y =-0.04764+0.23455 X+0.08031 X
0.0 0.5 1.0 1.5 2.0 2.5
RGA Detector Response (P^^ / P/100)
Figure A-1. Calibrated RGA Response for Argon
A-4
-------
(a)
9/11/07 Krypton Calibration
1.0-,
0.8-
o
§ 0.6-
O
CD
2
.Q
0.4-
0.2-
0.0-
Y =-0.01678+0.35308 X+0.03926 X"
0.0 0.5 1.0 1.5 2.0
RGA Detector Response (P^^ / P/100)
(b) 9/13/07 Krypton Calibration
2.5
1.0-.
0.8-
O
O 0.6-
O
£
"CD 0.4-
ca
O 0.2-1
0.0-
Y =-0.02032+0.26523 X+0.12842 X"
0.0
0.5
1.0
1.5
2.0
RGA Detector Response (Pm/e=84 / P/100)
Figure A-2. Calibrated RGA Response for Krypton
A-5
-------
Appendix B - Measurement Study Protocol
The analytical measurement and data interpretation approach described herein attempts to
determine, by empirical means, the most conservative cover gas destruction possible for a given
type of magnesium process tool and gas flow ranges used during production. This approach is
based on the experience from the MagReTech study and pertains to only those processes that can
allow typical operation and cover gas flow without molten metal in place, as well as normally
with molten metal. Therefore, ingot casting and chilling machines are prime candidates for this
approach, as opposed to alloying and die casting crucibles which are often kept under constant
high temperatures and filled with metal. The approach encompasses all the measurement
variance brought about by the process gas flows (including turbulence invoked by ambient air
dilution) in conjunction with analytical instrument and sampling variability. The variances
associated with each measurement condition needed in calculating the degradation factor are
then properly propagated through the calculations to the final result. The maximum destruction
factor is thus considered by adding the propagated variance to the final calculation result. If the
process under study is normally run over a range of cover gas concentrations, this measurement
approach is to be conducted at both the lowest concentration and highest concentration of that
range; the reportable maximum destruction factor is then the greater of the two.
The Test Plan outlines how the experimental observables and variances needed to
determine the maximum degradation factors are obtained. The Quantitative Data Analysis
section describes how the measurements are used to estimate destruction factors and how the
variances are propagated. A hypothetical example is provided to help illustrate the measurement
approach.
Test Plan
1. Set up a real-time measurement instrument to continuously extract and analyze a low
volume slipstream of the completely blended cover gas mixture prior to injection into the
melt protection area. The extractive analysis technique should not significantly impact
the overall gas flow within normal process operations, and allow enough consecutively
recorded measurements as to be statistically relevant. For instance, an extractive FTIR
system monitoring SFe concentrations over one-minute of signal averaging will produce
240 data points within a four-hour continuous sampling period.
B-l
-------
2. Set up a second real-time measurement instrument to continuously extract and analyze a
low volume stream of cover gas within the process environment (i.e., where molten
magnesium is to be covered during casting or chilling). The extractive analysis technique
at this sample location should not significantly impact the overall gas flow throughout the
process environment, should allow the most representative sampling configuration
possible (for instance, equidistant perforated sample probes that traverse the entire width
of process head space a few centimeters from the molten metal surfaces), and also allow
enough consecutively recorded measurements as to be statistically relevant. This
instrument will be sampling coincidentally with the instrument described in step #1, so
the implementation of two identical analysis techniques (for instance, two extractive
FTIR systems) is ideal.
3. With molten magnesium present during production: Simultaneously monitoring both
sampling locations over a sufficiently long time period will produce a mean concentration
during metal production at the point of cover gas injection, MP1, and a mean
concentration during metal production within the process environment, MPe. Also, the
99 percent confidence level of both means, which when expressing as +/- values about
the means would contain virtually all sources of indeterminate measurement error, can be
estimated as 2.58 times the standard deviations (divided by the square root of the number
of measurements) of their respective data sets. Hence, the experimentally determined
MP1/ MPe would carry associated OMP'/ OMP£ as total measurement uncertainties (the
squares of which being measurement variances) to be considered when calculating
degradation percentages.
4. With molten magnesium not present during mock production: Simultaneously monitoring
both sampling locations under the same configurations and over a similar length of time
as conducted in step #3 will produce, under nonmetal process conditions, an
experimentally determined NP1/ NPe pair and associated ONP'/ ONP£ measurement
uncertainties. These are needed to effectively estimate the amount of ambient air dilution
present in an open process.
5. Steps #3 and #4 are to be repeated as necessary over the cover gas concentration ranges
utilized during normal production, presumably since the degradations may be
significantly different depending on mixture ratios. This typically means an experimental
B-2
-------
pair will be run at the lowest and the highest operating concentration, with the highest
destruction percentage reported to be conservative.
6. Standard analytical measurement protocols, pertaining to the technique of choice, must
be run to characterize instrumental accuracy and reproducibility for each experiment (in
effect, any determinate errors associated with the instruments are quantified). Typically,
this means a pre-test and post-test calibration run for all instruments to ensure accurate
and consistent measurements at both sampling locations for each metal/nonmetal test
condition.
Quantitative Data Analysis
• For the noncasting sampling period, the reduction of injected cover gas concentrations
represents only the ambient air dilution in the system because molten magnesium is not
present to react with the cover gas. Therefore,
NPe
Dilution = 1 --
NP1
• For the casting or metal present sampling period, the reduction of injected cover gas
concentrations represents the ambient air dilution plus destruction in the system, for
molten magnesium is present. Therefore,
MPe
[Degradation + Dilution] = 1 -- -
• To determine solely the destruction factor, DF:
NPe MPe
DF = [Degradation + Dilution] - Dilution = - : --- - (1 )
NP' MP
• Since the sum of relative variances for each ratio yields the relative variance of the result
for each ratio in (1), and the sum of absolute variances for each term in (1) yields the
absolute variance of the difference of terms12, the variances of equation (1) can be
propagated and simplified to produce the variance associated with the destruction factor:
DF Np MP
Hence, the measurement uncertainty for DF is:
B-3
-------
1/2
(2)
Example
The noncasting sampling period produced mean SFe concentrations of 1000 ppmv and
500 ppmv for the cover gas injection point and process environment, respectively. The 99
percent confidence limit (basically, a multiple of the standard deviation) was 50 ppmv for each.
The subsequent metal sampling period produced mean SF6 concentrations of 1000 ppmv and 400
ppmv for the cover gas injection point and process environment, respectively. The 99 percent
confidence limit was also 50 ppmv for each. To summarize,
NP1 = 1000 ppmv;
------- |