LMSC-D406484
JUNE 30, 1975
STUDY OF
FACTORS AFFECTING REACTIONS
IN ENVIRONMENTAL CHAMBERS
FINAL REPORT - PHASE III
by
Raphael J. Jaffe
Prepared for
»
COORDINATING RESEARCH COUNCIL
NEW YORK, N.Y. • CONTRACTS CAPA 1-69 (1-72) AND (1-73)
ENVIRONMENTAL PROTECTION AGENCY
RESEARCH TRIANGLE PARK, NORTH CAROLINA • CONTRACTS 68-02-0287 AND 68-02-1270
-------
LMSC-D406484
June 30, 1975
STUDY OF
FACTORS AFFECTING REACTIONS
IN ENVIRONMENTAL CHAMBERS
Final Report
by Raphael J. Jaffe
Prepared for
COORDINATING RESEARCH COUNCIL
New York, N.Y.
Under Contracts CAPA 1-69(1-72) and (1-73)
and
ENVIRONMENTAL PROTECTION AGENCY
Research Triangle Park, North Carolina
Under Contract 68-02-0287 and 68-02-1270
Biotechnology
Lockheed Missiles & Space Company, Inc.
Sunnyvale, California
LOCKHEED MISSILES & SPACE COMPANY. INC.
-------
LMSC-D406484
PROJECT PERSONNEL
Name
Raphael J. Jaffe
Frank C. Smith, Jr.
Ken W. Last
Michael W. Reeder
E. H. Kawasaki
R. C. Tuttle
Dr. H. S. Johnston, Consultant
Area of Contribution
Project Direction
Analytical Chemistry/Chamber Operation
Statistical Analysis
Statistical Analysis
Analytical Chemistry
Analytical Chemistry
Photochemi stry
PROJECT MONITORSHIP
COORDINATING RESEARCH COUNCIL
PROJECT CAPA 1-69
Member
Mr. D. B. Wimmer, Chairman
Mr. Frank Bonamassa
Mr. Basil Dimitriades
Dr. J. M. Heuss
Mr. Stanley Kopezynski*
Dr. Hiromi Niki
Dr. E. E. Wigg
Affiliation
Phillips Petroleum Company
California Air Resources Board
Environmental Protection Agency
General Motors Corporation
Environmental Protection Agency
Ford Motor Company
Esso Research & Engineering Company
*Until January 1973
ii
LOCKHEED MISSILES & SPACE COMPANY, INC.
-------
LMSC-D406484
SUMMARY
An experimental study has been conducted of effects of materials, spectrum, surface/
volume ratio (S/V) and cleaning technique on the photochemical reactions observed in
a smog chamber. A unique chamber and lighting system was used, which permitted
independent variation in chamber materials and in light conditions. A xenon arc lamp-
parabolic reflector combination provided a collimated light beam. By orienting plates
of materials parallel to the beam, it has been possible to independently vary light con-
ditions and materials.
The propylene (3 ppm)/NO (1.5 ppm) reaction system was used, at 95° F and 25%
J\.
relative humidity. Initial NO0 content was nominally 10% of NO . Chamber background
x -1
was <0.1 ppm C. All photochemical runs were at k, ofO.Smin as determined by
frequent NCL in N_ photolysis tests.
The study included four materials — Teflon, Pyrex, aluminum and stainless steel, and
two conditions each of spectrum, S/V, and cleaning. A complete factorial testing
sequence was performed, with many replicates. Effects of the four materials and the
two levels of each parameter have been determined. The base chamber was mainly
Pyrex at an S/V of 1.4 ft" . The four materials were added to the chamber at S/V of
1.3 and 2.7 ft" . The full spectrum extends from 300 nm to higher wavelengths, simu-
lating sunlight at the earth's surface. The cut spectrum starts at 340 nm. .Two cleaning
methods - vacuum offgassing or purging at 110° F - were used.
The table below summarizes the results by giving the data averaged in several ways
for each material. For each material, eight conditions (2 spectrum x 2 S/V x 2
cleaning) were used and the "all runs" average gives the average result. "Full
Spectrum" average gives the average of the four (2 S/V x 2 Cleaning) conditions at full
spectrum; and "Cut Spectrum" gives the average of the four conditions at cut spectrum.
High S/V-Low S/V and Vacuum Clean-Purge Clean averages have similar meanings.
The last entry-Full, Low S/V, Purge gives the results for that specific condition.
iii
LOCKHEED MISSILES & SPACE COMPANY. INC.
-------
LMSC-D406484
Of the three factors studied, spectrum is the factor that has the largest effect. The
presence of this large spectral effect at constant light intensity - as measured by
constant k, -was not anticipated. However, it has been explained by consideration
of the spectrally sensitive photochemical reactions, and in fact appears upon reinter-
preting old literature data. The cut-off spectrum lowers the photochemical reaction
rate by about 50%, as compared to the full spectrum. Chamber size (S/V) is the next
most important factor, and the larger chamber usually shows a higher oxident maxi-
mum and a slower reaction. Cleaning technique is important for maximum oxidant
(vacuum cleaning gives higher values) and in the behavior of stainless steel.
The comparison between materials shows that stainless steel behaves in a manner
unlike the other three materials. The photochemical reaction is slowest in the presence
of Teflon, with increasing reaction rates for Pyrex, aluminum, and stainless steel.
Tests were also conducted at lower relative humidity. For stainless steel and aluminum,
lowering the relative humidity lowered the reaction rate. For Teflon, the opposite
effect occurred.
Several runs were conducted for the n-butane (3 ppm)/NO (0.6 ppm) system. Results
X
for the full spectrum conditions were similar to those previously reported in the
literature.
The ultimate goal of this project was to determine how various design and operational
factors affect reactions in smog chambers, in an attempt to reconcile the observations
reported from some ten smog chambers previously involved in a round-robin test
program. Regression equations have been determined that account for 75 to 90% of
the observed variability. These models use the following independent variables:
chamber material S/V, chamber light intensity (k,), spectral distribution correction
factor, chamber volume, and initial conditions.
iv
LOCKHEED MISSILES & SPACE COMPANY. INC.
-------
LMSC-D406484
RESUUPS SUMMARY
Material
Teflon Pyrex Aluminum Stainless Steel
N02 FORMATION RATE (ppb/min)
Averages
All runs 8.63 10.0 13.1 25.4
Full Spectrum 10.2 12.6 15.3 30.2
Cut Spectrum 7.4 7.4 10.9 20.6
High S/V 8.50 9.7 13-3 32.2
Low S/V 8.76 10.4 12.9 18.6
Vacuum Clean 7.77 10.2 12.4 20.2
Purge Clean . 9.58 9-7 13-8 30.5
Full, Low S/V, Purge 11.1 10.1 l8.i 25.6
TIME TO N0p MAXIMUM
All runs 154 131 106 70
Full Spectrum Il6 94 86 56
Cut Spectrum 192 168 126 84
High S/V 144 150 106 59
Low S/V 164 112 106 81
Vacuum Clean 178 129 118 80
Purge Clean l62 132 110 60
Full, Low. S/V, Purge 115 102 76 6r
TIME TO 50$ PROPYLENE CONSUMPTION (min)
An runs l80 157 133 97 .
Full Spectrum ' l45 179 111 80
Cut Spectrum 215 135 15^ Il4
High S/V 172 119 135 86
Low S/V 187 195 131 108
Vacuum Clean 183 153 139 8j
Purge Clean 177 . l6l 127 107
Full, Low S/V, Purge l44 120 92 8l
LOCKHEED MISSILES & SPACE COMPANY. INC.
-------
LMSC-D406484
Teflon Pyrex Aluminum Stainless Steel
MAXIMUM OXIDANT CONCENTRATION £ppm)
All runs 1.03 -93 1-00 .90
Full Spectrum 1.13 -98 1.02 1.10
Cut Spectrum .93 _.89 «98 .70
High S/V . 1.00 .8? .93 -97
Low S/V 1.06 .99 1-07 1-03
Vacuum Clean 1.11 1.00 l.Cfc .92
Purge Clean .96 .87 .96 .88
Full, Low S/V, Purge 1.11 .98 l.C4 1.1^
vi
LOCKHEED MISSILES & SPACE COMPANY. INC.
-------
LMSC-D406484
CONTENTS
Section Page
PROJECT PERSONNEL ii
SUMMARY iii
ILLUSTRATIONS xi
TABLES xiii
1 INTRODUCTION 1-1
2 EXPERIMENTAL METHOD 2-1
2.1 Test Conditions 2-1
2.2 Apparatus 2-2
2.2.1 Irradiation Chamber 2-2
2.2.2 Illuminator 2-8
2.2.3 Thermal Enclosure 2-9
2.2.4 Material and S/V Changes 2-11
2.2.5 Spectral Distribution and Spectral Changes 2-13
2.2.6 Cleaning Technique 2-15
2.2.7 Chamber Charging Technique 2-16
2.3 Chemical Analysis Methodology 2-16
2.3.1 NO0-NO 2-16
ti X
2.3.2 Ozone 2-19
2.3.3 Total Hydrocarbons 2-21
2.3.4 Propylene 2-22
2.3.5 Acetaldehyde 2-22
2.3.6 Peroxyacetyl Nitrate (PAN) 2-22
2.3.7 Moisture (Water) 2-23
3 RESULTS 3-1
3.1 Detailed Statistical Analysis 3-10
3.1.1 Covariate Analysis 3-10
3.1.2 Principal Components 3-10
vii
LOCKHEED MISSILES & SPACE COMPANY. INC.
-------
LMSC-D406484
Section
Appendix
A
3.1.3 Effects by Materials
3.2 Material Difference s
3.3 Effects of Factors
3.4 Ozone Decay Results
3.5 Relative Humidity Effects
3.6 Butane Runs
3.7 Background Reactivity Runs
3.8 "Virgin Surface" Effect
3.9 Nitrogen Balance
DISCUSSION
4.1 General Observations
4.2 NO2 Photolysis Distribution
4.3 Possible Mechanisms
4.4 Background Reactivity Runs
4.5 Ozone Decay Results
RECONCILIATION OF CAPI-6 DATA
5.1 CAPI-6 Data Handling
Factorial Experiment Data Handling
Multiple Regression Results
5.2
5.3
5.4
Normalization for k, Effects
d
5.5 Computations Using Normalized CAPI Data
5.6 Correlation Coefficients
RECOMMENDATIONS
REFERENCES
PHOTOCHEMICAL RUN DATA
A.I Propylene Graphs
A. 2 Propylene Data Tabulation
STATISTICAL ANALYSIS
1.0 Introduction and Overview
2.0 Background
2.1 The Experimental Design
Page
3-11
3-16
3-16
3-18
3-19
3-22
3-24
3-24
3-25
4-1
4-1
4-3
4-4
4-6
4-7
5-1
5-4
5-4
5-5
5-13
5-16
5-16
6-1
R-l
A-l
A-l
A-l
B-l
B-l
B-3
B-4
vni
LOCKHEED MISSILES & SPACE COMPANY. INC.
-------
LMSC-D406484
Section Page
2.2 The Data B-4
2.3 Covariates B-4
2.4 The Unbalanced Design and Its Consequences B-7
2.5 Principal Comments B-8
2.6 Outliers B-8
2.7 Time Trends B-10
2.7.1 Drifts within Material B-15
2.7.2 Material to Material Time Drift Analysis B-20
2.8 Computer Program B-23
3.0 C ovarianc e Analysi s B - 24
3.1 Univariate Covariance Results B-24
3.2 Multivariate Covariance Results B-26
3.3 Comparison of Covariate Corrected Data to
Adjusted Data B-29
4.0 Analysis of Variance B-29
4.1 Results of Univariate Analysis B-32
4.2 Results of Multivariate Analysis B-39
5.0 Principal Components B-46
5.1 Selection B-46
5.2 . Calculation B-51
5.2.1 Time B-53
5.2.2 Dose B-53
5.2.3 MISC I and MISC II B-54
5.3 Results and Interpretation B-55
5.3.1 Stainless Steel B-55
5.3.2 Teflon B-73
5.3.3 Pyrex B-81
5.3.4 Aluminum Reruns B-88
5.4 Univariate Analysis B-96
5.5 Results of Multivariate Analysis B-102
5.6 Recommended Model B-102
ix
LOCKHEED MISSILES & SPACE COMPANY. INC.
-------
LMSC-D406484
ILLUSTRATIONS
Figure Page
2-1 Smog Chamber Assembly 2-4
2-2 Stand Assembly 2-6
2-3 Environmental Chamber Showing Side Stream Mixer, Charge
Ports, and Clean-Out Port 2-7
2-4 Chamber Inside Thermal Enclosure 2-10
2-5 Arrangement of Surface Plates 2-12
2-6 Measured Spectral Irradiance Inside LMSC Smog Chamber -
Full and Cut Spectra 2-14
2-7 Typical Raw Data for Determining k, 2-15
2-8 Smog Chamber During Vacuum Off-Gassing Cleaning 2-17
2-9 Typical Linearity Check of NO Instrument 2-20
3-1 Composite Photochemical Test Results for Teflon Film Surfaces 3-5
3-2 Composite Photochemical Test Results for Pyrex Surfaces 3-5
3-3 Composite Photochemical Test Results for Aluminum Surfaces 3-5
3-4 Composite Photochemical Test Results for Stainless Steel Surfaces 3-5
4-1 Distribution of NO2 Photodisintegrations for Various Spectra 4-3
B5-1 Time to NO2 Maximum vs NOg Formation Rate B-50
B5-2 Dose vs. Run Sequence B-63
B5-3 Time vs Run Sequence B-63
B5-4 Misc I vs Run Sequence B-65
B5-5 Misc II vs Run Sequence B-65
B5-6 Stainless Steel Dose vs Time B-67
B5-7 Stainless Steel Misc I vs Time B-68
B5-8 Stainless Steel Misc II vs Time B-69
B5-9 Stainless Steel Dose vs Misc I B-70
B5-10 Stainless Steel Dose vs Misc II B-71
B5-11 Stainless Steel Misc I vs Misc II B-72
LOCKHEED MISSILES & SPACE COMPANY. INC.
-------
LMSC-D406484
Figure Page
B5-12 Teflon Dose vs Time B-75
B5-13 Teflon Misc I vs Time B-76
B5-14 Teflon Misc H vs Time B-77
B5-15 Teflon Dose vs Misc I B-78
B5-16 Teflon Dose vs Misc II B-79
B5-17 Teflon Misc I vs Misc II B-80
B5-18 Pyrex Dose vs Time B-82
B5-19 Pyrex Misc I vs Time B-83
B5-20 Pyrex Misc H vs Time B-84
B5-21 Pyrex Dose vs Misc I B-85
B5-22 Pyrex Dose vs Misc H B-86
B5-23 Pyrex Misc I vs Misc H B-87
B5-24 Aluminum Dose vs Time B-89
B5-25 Aluminum Misc I vs Time B-90
B5-26 Aluminum Misc II vs Time B-91
B5-27 Aluminum Dose vs Misc I B-92
B5-28 Aluminum Dose vs Misc II B-93
B5-29 Aluminum Misc I vs Misc n B-94
B5-30 Mean Values by Materials B-95
XI
LOCKHEED MISSILES & SPACE COMPANY. INC.
-------
LMSC-D406484
TABLES
Table Page
1-1 Characteristics of Chambers Used for Previous Inter comparison
Investigations 1-2
2-1 Chamber Description 2-8
3-1 Photochemical Test Calculated Parameter Definitions 3-2
3-2 Effects by Material 3-6
3-3 Effects by Material 3-12
3-4 Significant Effects by Materials 3-17
3-5 Maximum Ozone Concentrations (ppm) 3-17
3-6 Cut/Full Spectrum Ratio 3-16
3-7 Ozone Half-Life Study 3-18
3-8 Correlation of Ozone Half-Life with Run Parameters Averaged
Over Cleaning 3-20
3-9 Relative Humidity Variation Effects 3-21
3-10 Butane Effects by Materials 3-23
3-11 NO0 at Maximum/NO Initially 3-26
Ll X
3-12 NO2 Rate Divided by NO Rate 3-27
4-1 "Mylar/Teflon" Spectral Effect 4-2
4-2 Spectral Effect on Time to NO0 Maximum Caused by Each Species 4-6
£t
4-3 Background Reactivity Results • 4-7
5-1 Chamber Characteristics 5-2
5-2 CRC-APRAC Irradiation Chamber Comparison 5-3
5-3 Results of Multiple Regression with Full Model 5-7
5-4 Run Parameters Predicted From Multiple Regression 5-8
5-5 Run Parameters Residuals from Multiple Regression 5-9
5-6 Normalized Residuals from Multiple Regression 5-10
2
5-7 Comparison of R from Dual and Single k. Models 5-11
5-8 Reduced Multiple Regression Models 5-12
xn
LOCKHEED MISSILES & SPACE COMPANY. INC.
-------
LMSC-D406484
Table Page
5-9 Effect of k. Variation 5-14
a
5-10 CAPI Chambers Adjusted to kd of 0.4 5-15
5-11 Correlation Coefficients of Run Parameters 5-17
5-12 Correlation Coefficients of Run Parameters and Ozone
Half-Lives 5-19
B2-1 Standard Deviation of Parameters B-6
B2-2 Rank Sum and Sign Test Results for Adjusted Data B-16
B2-3 Sign Test Results for Covariate Corrected Data B-17
B2-4 Covariate Corrected Data (Original Alum Omitted) B-18
B2-5 Means by Material B-21
B3-1 Univariate Covariate Adjustment Coefficient B-25
B3-2 Multivariate Covariate Adjustment Coefficients B-27
B3-3 Results of Covariate Multrivariate Forward Approach B-28
B3-4 Comparison of Covariate Corrected Data to Adjusted Data B-30
B4-1 Contrasts Used in the Analyses of Variance B-31
B4-2 Significance of Effects & Interactions on Univariate Data Given
HC Init and % NO2 B-33
B4-3 Univariate Prediction Coefficients for Covariate Adjusted Data B-34
B4-4 Significance of Univariate Material Contrasts Given HC Init
and % NO2 B-38
B4-5 Univariate Original Aluminum Analysis, After HC Init and
% N02 B-40
B4-6 Univariate Pyrex Analysis After HC Init and % NO2 B-41
B4-7 Univariate Teflon Analysis, After HC Init and % NO2 B-42
B4-8 Univariate Stainless Steel Analysis, After HC fait and % NO2 B-43
B4-9 Univariate Rerun Aluminum Analysis, After HC fait and % NO- B-44
B4-10 Abbreviated MANOVA, Given HC fait, TADJ, and NO B-45
' A.
B5-1 Correlation Coefficients for Unadjusted Data B-46
B5-2 Correlation Coefficient for Adjusted Data B-48
B5-3 Correction Coefficients for Covariate Adjusted Data B-49
B5-4 Ratio of Mean Variation to Error Variation B-52
B5-5 Transformations Used fa Obtaining Principal Components B-56
xiii
LOCKHEED MISSILES & SPACE COMPANY. INC.
-------
LMSC-D406484
Table Page
B5-6 Principal Components Values B-57
B5-7 Material Effects for Principal Components B-74
B5-8 Univariate Tests of Effects B-97
B5-9 Univariate Tests of Reduced Models B-98
B5-10 Optimum Univariate Models on Principal Components B-99
B5-11 Coefficients for Univariate Models After Prin Components B-100
B5-12 Material Contrasts in Optimum Models B-101
B5-13 By Materials B-103
B5-14 Tests of Reduced "By Materials" Models B-104
B5-15 "By Material" Prediction Coefficients for Univariate Principal
Components Data B-105
B5-16 Multivariate Test of Effects B-106
B5-17 Coefficients for Multivariate Model After Principal Components B-107
xiv
LOCKHEED MISSILES & SPACE COMPANY. INC.
-------
LMSC-D406484
Section 1
INTRODUCTION
Chambers in which systematic studies can be made of the reactions between
hydrocarbons and nitrogen oxide in the presence of simulated sunlight have
been in use for some twenty years. Such chambers have generally been
successful in simulating the gross features of photochemical smog, such as
production of oxidants and eye irritants, and haze. Intercomparison of the
results obtained in these smog chambers has not been extensively attempted
until relatively recently (Ref. 1), at which time the interest in individual
hydrocarbon reactivity measurements led to an understanding of the need to
compare the various facilities. Intensive comparisons of results obtained in
ten smog chambers have been performed by the Coordinating Research
Council project CAPI-6, Techniques for Irradiation Chamber Studies, and
CAPA 1-69 (Factors Affecting Reactions in Environmental Chambers). The
range of physical characteristics of these chambers is shown in Table 1-1.
A group of round-robin tests was conducted using these chambers as follows:
(1) irradiation of seven different hydrocarbons with nitrogen oxide; (2)
replicate runs to establish reproducibility using the propylene-nitrogen oxide
system; (3) a reactant concentration study in which 3. 0 ppm propylene was
reacted with 3. 0, 1. 5, and 0. 5 ppm nitrogen oxides (Ref. 2).
The differences observed among the chambers could not be accounted for
analytically, and an experimental study of how various design and operational
variables affect the photochemical reactions observed in smog chambers was
instituted. This is a report of the results of this study, which are also the
first reported results for a smog chamber illuminated by a xenon arc lamp.
The facility was developed and initial tests performed under Phase I of the
project, which has been previously reported (Ref. 3). This report covers
Phase II and III of the project, and includes all information contained in the
Phase II report (Ref. 4).
1-1
LOCKHEED MISSILES & SPACE COMPANY, INC.
-------
LMSC-D406484
Table 1-1
CHARACTERISTICS OF CHAMBERS USED FOR PREVIOUS
INTERCOMPARISON INVESTIGATIONS
Volume (ft0)
Surface/Volume Ratio (ft
Surface Type as S/V
Stainless Steel
Aluminum
Glass
Plastic Film
Light Intensity
(kd, rain"1)
Type of Lighting
-1.
2.9 to 1140
0.78 to 4. 91
0 to 2.44
0 to 0.92
0 to 2.81
0 to 0.83
0.16 to 0.6
Fluorescent lamp combinations, of
sunlamps, black lamps, and blue
lamps (both internal and external)
1-2
LOCKHEED MISSILES & SPACE COMPANY. INC.
-------
LMSC-D406484
Section 2
EXPERIMENTAL METHOD
2. 1 TEST CONDITIONS
The following combinations of factors have been studied in a full factorial test
plan:
Materials S/V Spectrum Cleaning
Aluminum 2.7 ft" Full Vacuum Off-gas
Pyrex 1.3ft'1 Cutoff Purge at 110°F
Teflon
Stainless Steel
The testing sequence consisted of performing the photochemical tests for the
aluminum surfaces, followed by Pyrex, Teflon, stainless steel, and a retest
of the aluminum. This has allowed an analysis for time trend, to see if a
systematic drift was present in the experiment. A number of replicate tests
were performed. These were distributed among immediate replicates,
replicates with 1 to 15 intervening runs, and the aluminum re-test replicates,
which had greater than 50 intervening runs.
The propylene/NOx system was used for most tests, at 3. 0 ppm propylene and
1. 5 ppm NOX. The initial NO£ content was nominally held at 10 percent. The
chamber was held at 95±3°F throughout all tests. Relative humidity was 25 t
5 percent (49 to 59°F dew point). Chamber pressure was slightly above
atmospheric (0. 1 in. H^O). Zero air was used to maintain chamber pressure,
to make up for sampling and leakage, at about 3 percent/hour make-up rate.
A series of tests were performed at lower relative humidity. Another set of
tests were performed for the n-butane 3. 0 ppm)/NOx(0.6 ppm) system.
2-1
LOCKHEED MISSILES & SPACE COMPANY. INC.
-------
LMSC-D406484
2.2 APPARATUS
The apparatus used was specifically designed to meet the objectives of the study by
allowing independent variation in materials configuration, lighting conditions, and
cleaning technique. The illuminator produces a collimated light beam. Plates of
materials can be placed parallel to the beam axis without affecting the light conditions.
This decoupling of light and materials allows independent variation of the two factors.
Major apparatus items are the smog chamber, the xenon arc lamp illuminator, the
thermal enclosure for the chamber, and the gas analysis instrumentation.
2.2.1 Irradiation Chamber
The irradiation chamber is hexagonal in cross section, measuring 54 in. across the
diagonal of the cross section. The chamber configuration is shown in Fig. 2-1 and the
chamber support stand is shown in Fig. 2-2. Figure 2-3 represents a pictorial view
of the chamber. The chamber is constructed of six flat side panels that fit into an
extruded aluminum framework. The aluminum framework is coated with 5 to 8 mils
of FEP Teflon. The resultant panels are bolted together and are supported from cir-
cular rings on the stand.
The faces of the chamber are of tempered 1/4 in. Pyrex glass to admit the light and
pass the beam through the chamber with minimal reflection or absorption. The flat
sides of the chamber are also fabricated from 1/4 in. tempered Pyrex glass. The
working stress of the tempered Pyrex is 3600 psi, which results in an allowable pressure
differential of 7.1 in. of water. This is based on the hexagonal faces, which are the
weakest members. Sealing of the 1/4 in. Pyrex to the aluminum extrusion is accom-
plished with silicon rubber gaskets that have been off-gassed at about 10 torr for
over 24 hours. The gasket is fitted between the aluminum channel and the Pyrex, and
the Pyrex is pressed into a channel section in the extrusion. A silicon rubber O-ring
is used to seal adjacent aluminum extrusions which are bolted together. Additional
sealing is accomplished with a coat of Dow Corning 30-121 RTV silicon rubber on all
external joints.
2-2
LOCKHEED MISSILES & SPACE COMPANY. INC.
-------
LMSC-D406484
Three 1-in. diameter ports are located in one of the lower side panels. The ports are
fabricated from brass and are coated with FEP Teflon. Sealing of these ports to the
Pyrex chamber walls is accomplished with silicon rubber Baskets. Sealing of tubes
inserted into the ports is accomplished'with Teflon seals. Tubes can be fitted into
these ports and adjusted to various distances into the hexagon. The rear port is used
for a gas thermocouple, which is Teflon coated and shielded from direct illuminator
radiation. The front port has a 1-in. diameter Pyrex tube installed that is used for
venting the chamber.
The center port is used for gas sampling. A 1-in. diameter Pyrex tube containing a
concentric 7-mm tube is used. The sample enters at about the geometric center of
the chamber, and is drawn to a sample manifold through a 1/4-in. Teflon tube, and
thence to the analysis instrumentation. The sample contacts only Pyrex or Teflon.
before entering the instruments. Delay time from sample withdrawal to instrument
inlet is about 30 seconds. The effect of this delay is discussed in Section 3.9.
A cleanout port is located in this same Pyrex panel. This port is 3-in. diameter and
is used for chamber outgassing during vacuum off-gas cleaning. A Pyrex disc with a
silicon rubber gasket is used to seal this port during normal operation. Metallic parts
are all coated with FEP Teflon. A Pyrex relief valve is also mounted on this panel.
This relief valve has a 2-in. diameter opening leading to the chamber. The valve
utilizes a water seal principle and the relief setting is adjusted by the addition or
withdrawal of water. Both negative and positive differential pressure protection are
provided. An overflow trap is included to prevent water from entering the chamber.
A side stream mixer is also located on this panel. This mixer consists of a 6-in.
diameter duct connected to the panel at a point 1/3 of the way from the front and 1/3
of the way from the rear of the chamber. Gas is circulated through the duct by means
of a Teflon-coated fan blade installed in the duct. The blade is connected to a shaft
that penetrates the duct through a rotary Teflon seal. The shaft is connected to a
motor via a pulley and belt which are located outside the chamber. The fan is rotated
2-3
LOCKHEED MISSILES & SPACE COMPANY. INC.
-------
Fig. 2-1 Smog Chamber Assembly
2-4
LOCKHEED MISSILES & SPACE COMPANY. INC.
-------
IMSC -
Sf/00-07
-------
o
n
m
m
o
U)
U)
F
m
U)
fi?
o
m
o
0
2
TJ
Z
o
to
O5
Fig. 2-2 Stand Assembly
£
o
£
g
-------
LMSC-D406484
Fig. 2-3 Environmental Chamber Showing Side Stream Mixer,
Charge Ports, and Clean-Out Port
2-7
LOCKHEED MISSILES & SPACE COMPANY. INC.
-------
LMSC-D406484
at 250 rpm, producing a circulation rate of approximately 75 cfm. The inlet end of
the duct enters the chamber at a 45-deg angle to minimize short circuiting of the gas
circulation. All parts of the side stream recirculator are coated with Teflon. The
mixer is operated during charging of the chamber only. It is shut off during photo-
chemical tests.
The chamber is mounted at each hexagonal apex to a circular structural framework,
which in turn mounts to a dolly with casters for easy transport of the chamber. The
structural frame is configured to allow chamber assembly within the framework. A
locating jack system is used to adjust the height of the chamber to exactly match the
height of the light source.
Geometric characteristics of the chamber are summarized in Table 2-1.
Table 2-1
CHAMBER DESCRIPTION
Configuration Hexagonal Prism
Length 60 in.
Diagonal 54 in.
Volume 65.9ft3
Surface Area
End Plates 26.4ft2
Side Plates 67.5ft2
Total 93.9ft2
S/V . 1.43ft'1
2.2.2 Illuminator
The irradiation source for the chamber is external to the chamber. This external
source consists of an arc lamp situated in front of a large parabolic reflector.
2-8
LOCKHEED MISSILES & SPACE COMPANY. INC.
-------
LMSC-D406484
Collimated energy from the reflector is directed toward the chamber. The collimated
beam used as the light source for the irradiation chamber is 5 ft in diameter. The
light source for the illumination system is an air cooled Osram compact xenon arc
lamp (6,500 watt nominal rating). The fireball for the lamp is 2.4 by 9 mm. The
optical system consists of a 5-ft diameter parabolic primary mirror and a spherical
back-collector mirror. The primary mirror is made of copper with a rhodium plating
with vacuum deposited aluminum over the rhodium. An SiO coating is used to protect
the aluminum. Reflectance is approximately 0.9 at a wavelength of 300 nm. The
spherical back reflector is similarly coated. The back reflector is utilized to capture
energy from the lamp that would normally not strike the parabola, and focus it back on
the parabolic reflector. The optical system collects about 35 percent of the lamp
radiated light and directs it as a collimated beam into the chamber front face. The
e\
lampholder casts a shadow about 1.5 ft in area, which obscures about 8 percent of
the beam. As the chamber gases are well mixed, and all samples are taken from one
representative lighted area of the chamber, the shadowed area of the chamber has no
appreciable effect on the data.
For the cut spectrum configuration, a plane reflector is mounted behind the chamber
rear face. This reflector provides a second pass of the light through the chamber. It
is a front-surfaced, aluminized SiO -coated reflector, to maintain reflection down to
X
300 nm wavelength.
The illuminator is integrated into a searchlight housing. The housing is on casters,
which allows the illuminator to be moved readily.
2.2.3 Thermal Enclosure
A thermal enclosure is used to control the chamber temperature during a photochemical
run. Design requirements for the thermal enclosure are maintenance of chamber gas
temperature at 95 ± 3°F. The enclosure, depicted pictorally in Fig. 2-4 with the
chamber installed, consists of a plywood housing insulated with fiberglass, with heated
air circulating throughout the enclosure. The enclosure has a hexagonal-shaped hole
in the rear to allow the chamber to protrude. The front end of the thermal enclosure
2-9
LOCKHEED MISSILES & SPACE COMPANY. INC.
-------
LMSC-D406484
Fig. 2-4 Chamber Inside Thermal Enclosure
2-10
LOCKHEED MISSILES & SPACE COMPANY. INC.
-------
LMSC-D406484
has sliding doors that allow easy removal and installation of the chamber. These doors
also have a hexagonal-shaped hole that will allow the forward end of the irradiation
chamber to protrude. This allows the illuminator energy to enter the chamber.
Both the rear and forward hexagonal openings in the enclosure fit tightly to the chamber
to minimize gas leakage from the enclosure. The interior surfaces of the enclosure
are insulated with 3 in. of fiberglass.
A centrifugal blower is utilized to circulate approximately 1500 cfm of air over four
1.3-kw heaters that are used for thermal control. Three of the heaters are manually s
switched on or off and the fourth heater is thermostatically controlled to maintain the
enclosure air temperature at a fixed level. This system allows use of the fixed heaters
for purge cleaning, or warm-up, and coarse temperature control, with fine temperature
control being accomplished with the thermostatically controlled heater. The thermal
sensor is a West resistance bulb controller. This controls the air temperature in the
thermal enclosure. Air circulation is accomplished with a centrifugal blower that
passes air through plenums located at the bottom of the thermal enclosure. The heaters
are located in these plenums. Air passes out of the plenums, over the chamber and
is withdrawn out of the top of the thermal enclosure, where it is recirculated to the
heater plenum. The thermal enclosure is approximately 8 ft high by 8 ft wide by 6 ft
long.
2. 2.4 Material and S/V Changes
Plates of the material under study are inserted into the chamber, oriented parallel
to the light-beam axis of collimation, and vertically. These plates vary from 5-3/8
to 22-3/8 in. in width and are the height of the chamber. Plates are arranged in six
sets, with each set running the length of the chamber. Frequent open spaces interrupt
each set of plates, and form openings that are the height of the chamber and 2 to
4-1/2 in. wide, to allow the sampling and vent tubes to penetrate into the chamber,
and to promote mixing of the entire chamber contents. Figure 2-5 gives the plate
layout. Either six or three sets of plates are used, which gives either 2. 7 or 1. 3 ft~
2-11
LOCKHEED MISSILES & SPACE COMPANY. INC.
-------
5 B :
o
i X SAMPLE PORT INlfT
m
m
U)
TJ
>
o
m
o
0
Z
_ — - 60"
z
n
NOTE: Plate sets B, C, E removed for
low S/V configuration
01
O
o
05
Fig. 2-5 Arrangement of Surface Plates
-------
LMSC-D406484
of the material under study. The metallic materials are polished to a mirror finish.
The materials details are aluminum — 1100 H 14 alloy, stainless steel — type 304,
Pyrex-Corning 7740 plates, and Teflon - 5 mil FEP film.
2.2.5 Spectral Distribution and Spectral Changes
The spectral interval of interest in atmospheric photochemical simulation is 300 to
about 400 nm. The lower wavelength is the natural cutoff provided by the earth's
ozone layer, and the upper wavelength is set by the energy required to dissociate
NO?. It is generally recognized that compact xenon arc lamps provide the best
available match for this UV region.
Two spectral distributions were used in these experiments. The full spectrum con-
figuration is shown in Fig. 2-6. The second distribution is the cutoff spectrum, and
is also shown in the figure. This is obtained by placing a sheet of 3/16 in. thick
Plexiglas between the light source and the smog chamber. To avoid aging effects, a
fresh sheet of Plexiglas is used after about each fifth test. Total light intensity is
restored to the same value as used for the full spectrum runs by providing a second
pass of the light through the chamber. The front surfaced aluminized reflector is
used for this purpose.
Spectral data were taken with an Optronics Laboratory spectroradiometer. The radi-
ometer consists of a calibrated photovoltaic cell, a grating prism spectrometer, and
-9 -2 2
blocking filters. Digital readout is provided over a range of 10 to 10 watt/cm -nm.
Bandpass is 5 nm. The unit is calibrated against an NBS standard quartz iodide lamp,
over the wavelength interval of 250 to 1100 nm. Note that the characteristic xenon arc
lamp peak at 467 nm is easily seen by the radiometer. These data are in agreement
with earlier data taken using a photomultiplier tube spectroradiometer unit.
Measurements of the light beam were made at the plane corresponding to the front
window of the chamber, and of the beam at the exit of the chamber, after passing
through the front and rear 1/4-in. Pyrex window. To measure spectral irradiance in-
side the chamber, the rear window is removed, and the spectroradiometer entrance
2-13
LOCKHEED MISSILES & SPACE COMPANY. INC.
-------
LMSC-D406484
5
u
£
120
100
80
M
20
FUU ,
I I t L/ I I I I . I I I I
I
I
/CUT
290
300
HO 400
WAVELENGTH
-------
LMSC-D406484
i
8
o
Fig. 2-7 Typical Raw Data for Determining k,
To maintain the k, constant throughout the study, lamp power was adjusted. Intensity
was determined each time a new material was placed in the chamber and each time S/V
was changed. Measurement of k, was performed for both the full spectrum and the cut
spectrum. On the average k, was determined each four runs (including replicates).
To guard against any unanticipated unusual drift in the lamp light output, an intensity
meter was used to monitor the light. No unexplained readings were observed. It was
thus concluded that the entire set of runs submitted for statistical analysis was performed
at 0.3 min" , within perhaps a 10% variation.
2.2.6 Cleaning Technique
The two cleaning techniques used are: purging at 110°F, or vacuum off-gassing. Purge
cleaning consists of holding the chamber overnight at a temperature of 110 ±3° F while
4 to 6 chamber volumes of pure air are purged through the chamber. For vacuum off-
gassing, the smog chamber is moved to an adjoining 18 by 18 by 36-ft vacuum chamber.
2-15
LOCKHEED MISSILES & SPACE COMPANY. INC.
-------
LMSC-D406484
The vacuum chamber is pumped by Roots Blowers and mechanical pumps which prevents
any back migration of pump oil. The chamber walls are 304 stainless steel, polished
to a No. 4 mill finish. The smog chamber is held at about 2 microns pressure (or less)
for at least 16 hours. During this time, the chamber is maintained at about 100° F.
After the off-gassing, the smog chamber is repressurized by bleeding charge gas into
the smog chamber while repressurizing the vacuum chamber. Figure 2-8 shows the
smog chamber inside the vacuum chamber.
2.2.7 Chamber Charging Technique
The chamber is charged with synthetic compressed air. The air is purchased from a
single vendor to a specification of <0.1 ppm total hydrocarbons (as methane), and <10
ppm hydrogen. The air is produced by combining nitrogen gasified from liquid nitrogen
and electrolytic oxygen. The oxygen is made by electrolysis of distilled water. It then
is further purified by catalytic combustion of the trace hydrogen, followed by molecular-
sieve drying of the oxygen. A pre-charge determination of total hydrocarbons verifies
the air purity. The chamber is charged through a stainless steel humidifier, packed with
Rashig rings and filled with triple distilled water. Starting dew point is adjusted to
54 ±5°F by dilution with the dry pure air.
Reactants are added to the chamber from stock cylinders of about 350 ppm in nitrogen.
The reactant blends are transferred using separate 1/2 liter sampling cylinders. The
transfer cylinder is pressurized at about 200 psi, placed in the charge manifold, and slow-
ly bled to the smog chamber. In this technique, the transfer cylinder pressure (which is
read to 0.5 psi) becomes the reproducibility limit for the initial reactant charge.
2.3 CHEMICAL ANALYSIS METHODOLOGY
2.3.1 NO0-NOV
& X
Nitrogen dioxide is monitored by the modified Saltzman-Lyskow wet chemical technique,
utilizing a continuous sampling Technicon Autoanalyzer unit. The NO0 absorbing solu-
£t
tion is made from 2.0-gm N-1-naphthylethylenediamine dihydrochloride, 100-gm sulfa-
nilic acid, 5-cc Kodak Photoflo, and 1 liter glacial acetic acid in 5 gallons of water.
The lifetime of the solution is greater than 1 month when stored in an aluminum foil
covered bottle and shielded from air exposure. The gas sampling system consists of
2-16
LOCKHEED MISSILES & SPACE COMPANY. INC.
-------
LMSC-D406484
Fig. 2-8 Smog Chamber During Vacuum Off-Gassing Cleaning
LOCKHEED MISSILES & SPACE COMPANY. INC.
-------
LMSC-D406484
two 15-cm by 2.4-mm I.D. (28 turn) glass mixing coils in series where the gas sample
stream contacts the absorbing solution, an accumulator/liquid-gas separator, a flow-
meter to measure sampling rate, a chain driven peristaltic pump using variable flow
fluoroelastomer tubes, a colorimeter with a 50-mm flow cell, and an extended range
recorder. Gas sampling rate is 150 cc/min and liquid sampling rate is 1.5 cc/min.
When new absorbing solution is prepared, a static calibration using NaNO2 solutions
)2/ml is performed. NO2
-3 -1
ranging from 1. 5 x 10 to 1. 5 x 10 /*! NO0/ml is performed. NO0 gas concentra-
tions are determined from the formula
N°2 =
where:
A = microliters NO2 gas per milliliter of liquid standards
_ (mg/liter NaNO,) (24.5 liters/mole)
(Mol wt. NaNO2) 0. 72 moles NaNOg/mole NO2
where 24.5 is the molar volume at 25° C and 760 torr and 0.72
is the Saltzman factor
B = flow rate of absorbing solution reagent (ml/min)
C = gas stream flow rate (liters/min)
D = column efficiency (expressed decimally)
Daily dynamic calibrations are performed using standard Metronics 4-cm constant
rate NO2 permeation tube. The permeation tube is placed inside a constant-temperature
condenser. Low NO2 concentrations are obtained by passing compressed air through a
calibrated flow meter and then sweeping the NO2 from the permeation tube into a 12.5
liter dilution flask. The NO2 stream enters the bottom of the flask and sampling is done
at the top using a single 4-way path to the gas stream. The dilution flask pressure is
measured with a -0.5 to +0.5 in. water Magnihelic gage and is adjusted with a vent line
constriction to get pressure resembling chamber run conditions (+0.02 in. HgO). The
dilution system is all Teflon and glass except for small Tygon connections. The permea-j
tion tubes are calibrated gravimetrically and are within ±5% of the stated rates.
2-18
LOCKHEED MISSILES & SPACE COMPANY. INC.
-------
LMSC-D406484
Permeation tube NO2 concentrations are determined from the formula
(L) (K) (P )
C = = x 100
where:
C = gas concentration (pphm)
L = length of permeation tube (cm)
K = molar volume/molecular weight = 0.532 for NO2 at 25° C and 760 torr
(liter/gm)
P. = permeation rate of NO0 at temperature t (ng/min per cm of tube length)
3
F = carrier gas flow rate past tube (cm /min)
Nitric oxide is continuously measured with a Thermo Electron Corporation Model 12A
chemiluminescent gas monitor and Honeywell 18 recorder. This instrument measures
the chemiluminescent reaction of NO and Oq. Gas flow to the instrument is 150 cc/min.
The NO mode of the instrument uses a stainless steel converter run at 800° C to reduce
li
NO2 to measurable NO. Converter efficiency has been established at 99+percent. In-
strument zero, full scale and photomultiplier tube dark current are checked prior to
turning on the Oq generator. The instrument is calibrated daily by the dynamic NO,,
dilution gases and by a stock 88 ppm NO in No standard gas. This stock gas was checked
at 88 ppm by the supplier using long-path IR and a chemiluminescent monitor; and by an
independent chemiluminescent monitor in an informal exchange. Linearity of the instru-
ment is periodically confirmed by the exponential dilution technique. Precision is within
5 percent. Figure 2-9 shows an exponential dilution linearity check.
2.3.2 Ozone
Ozone is measured directly and continuously by a McMillan Model 1100 Ozone Meter.
This instrument measures the chemiluminescent reaction of ozone and ethylene. The
meter has four scales, with 0 to 2 ppm scale most commonly used. Daily meter cali-
bration is done at 1 ppm before and after a photochemical run by using a McMillan 1000
ozone generator. The output of the generator is regulated by sliding cover for the UV
lamp and output at 1.00 ppm (999 dial setting) is checked with a null meter. Periodic
2-19
LOCKHEED MISSILES & SPACE COMPANY. INC.
-------
100
80
60
40
20
10
i 8
Q_
£ 6
1'
a:
—
2
o
o
1
0.8
0.6
0.4
0.2
0
0
SCALE CHANGE
SCALE CHANGE
5
20
10 15
TIMEtmin.)
Fig. 2-9 Typical Linearity Check of NO Instrument
25
2-20
LOCKHEED MISSILES & SPACE COMPANY- INC.
-------
LMSC-D406484
calibrations of the generator and meter are done with the neutral buffered KI technique.
The 2-percent KI absorbing solution is calibrated with a stock !„ solution titrated
against standard Na0S0O0. The absorbance of the resulting KI/I0 solution is measured
£1 u G £
on a Perkin-Elmer 202 spectrometer at 350 run. Precision is about 4 percent. In the
first 20 months of usage of the ozone generator, the calibrations against neutral KI
showed no significant drift (< 5%). The null meter-front panel adjustment potentiometer
technique, which is used to maintain the ozone generating lamp light output, was
evidently effective. At that time the generator malfunctioned and was returned for
repair and recalibration. The good stability observed may also reflect the constant
air flow system and the air pre-treatment. Room air is filtered through an MSA Type
N canister and a pre-ozonation UV lamp. This lamp produces ozone which converts
trace NO to NO0. The air then goes through a molecular sieve 13X filter which destroys
Lt
the ozone. This scheme avoids loss of part of the calibrated ozone output by the
NO-OQ reaction.
o
Ozone concentration is corrected by checking 1. 00 ppm before and after a run. If a
discrepancy is noted, the initial and final readings at 1 ppm from the generator are
plotted linearly against time. Corrections are read from the line and applied to the
readings. Corrections are applied if the meter is more than 5 percent different from
the generator. The largest drift noted was about 18 percent.
2.3.3 Total Hydrocarbons
Chamber total hydrocarbons as CH4 are monitored with a F&M Model 700 gas chroma-
tograph with a Model 810 electrometer using an O0-H0 flame ionization detector and
£t £i
an unpacked 1/8 in. O.D. stainless steel column. Calibration is done from a zero air
cylinder with a THC concentration of 0. 07 ± 0. 03 CH.. The cylinder concentration
was determined at the supplier and checked upon delivery. Chamber samples from the
glass manifold are drawn through a 1/8 in. O.D. Teflon line and then through a sliding
sampling valve by a small vacuum pump.
2-21
LOCKHEED MISSILES & SPACE COMPANY. INC.
-------
LMSC-D406484
2.3.4 Propylene
Propylene is measured using the same detector and valve used for total hydrocarbons.
A Carle Microvolume Switching Valve is used for analysis selection.. The column for
propylene is .6 ft x 1/8 in. 100/120 mesh Porapak S stainless steel. The oven
temperature is 105° C; the detector is set at 150°; and the injection port is set at 125°.
The sampling loop has a 5. 0 cc volume, which gives an LLD of 0. 03 ppm.
Calibration was done using a stock cylinder. The concentration and linearity from 0.05
to 3. 00 ppm of the analysis were confirmed. Precision at 3.00ppm is less than 5 percent.
2.3.5 Acetaldehyde
Acetaldehyde uses the other detector of Model 700 gas chromatograph and a separate
Model 810 electrometer. The column is a 20-percent FFAP Chromosorb W (DMSC
treated), 60/80 mesh, 20 ft x 1/8 in. stainless steel column. The 10. 0 cc sampling
loop and valve use the same Teflon line from the manifold and vacuum pump as do the
propylene and total hydrocarbon analyses. Oven, detector, and injection port temper-
atures are the same as for propylene. Calibration is done with a stock cylinder and
linearity has been established from 0. 08 to 1.50 ppm, with an LLD of 0. 03 ppm and
precision less than 10 percent at 1.50 ppm. When the column is too noisy to give
useful data, the Porapak S column used for propylene can be used for acetaldehyde
analysis, although the acetaldehyde peak is broad and LLD is 0. 07 ppm. A comparison
of acetaldehyde concentration indicated by both columns shows quite good agreement.
2.3.6 Peroxyacetyl Nitrate (PAN)
Peroxyacetyl nitrate is measured on a Varian Model 600C gas chromatograph using a
tritium electron capture detector with an Ng carrier and a 5-percent Carbowax 600 on
60/100 mesh Chromosorb W (DMSC treated) 22 in. x 1/8 in. Teflon column run at
ambient temperature (25°C). The 0.5 cc sampling loop is made of Teflon and gives
2-22
LOCKHEED MISSILES & SPACE COMPANY, INC.
-------
LMSC-D406484
an LLD of about 0.003 ppm PAN. Standards are prepared by irradiation of an ethyl
nitrate-oxygen mixture with a UV lamp. Concentration of PAN in this standard is
determined by infrared analysis, and the 10 cm cell containing the mixture is purged
into a Tedlar bag with air measured by a wet test meter. Extensive calibration has
shown good linearity from 0. 05 to 1. 0 ppm with precision estimated at 25 percent.
2.3.7 Moisture (Water)
Moisture content was monitored with a Cambridge System Model 992 Hygrometer.
Samples were drawn periodically from the chamber manifold through a two-way
sampling valve. Otherwise, room air was purged through the instrument allowing
continuous instrument readout. Stable instrument response at 250 cc/min sample
flow was less than 3 minutes for dew points ranging from 48° F to 56° F. Precision
was less than 0.1 percent, based on saturation pressure.
2-23
LOCKHEED MISSILES & SPACE COMPANY. INC.
-------
LMSC-D406484
Section 3
RESULTS
Behavior of photochemical reaction systems is usually characterized by a
few chosen calculated parameters. In this work, the parameters used are
those defined by the CAPI-6-69 Project, with the addition of several measures
that appear interesting, for a total of 23, as defined in Table 3-1.
Parameter 17 was added to describe the NO disappearance as good NO data
are available from the chemiluminescent instrument.
Parameters 18 and 19 are calculated to give additional insight into the NO_
and oxidant dosage values, by normalizing them to a potential maximum
dosage represented by the denominator.
Parameter 20 describes the NO? curve to some extent, by giving the full
width at half maximum of the curve. It has been included to facilitate
comparisons of the NO? curve shape on a numerical basis.
Parameters 21 to 23 give various defined intervals in the photochemical run.
The crossover time is used because it is well demarked on the run graphs.
Measuring time from the crossover time yields system behavior characteristics
that are less dependent on initial conditions.
3-1
LOCKHEED MISSILES & SPACE COMPANY. INC.
-------
LMSC-D406484
Table 3-1
PHOTOCHEMICAL TEST CALCULATED PARAMETER DEFINITIONS
NO Formation
©
NO.
Rate = rTf; — where T- /0 is the time to form an amount of NO0 equal to 1/2
1/2 ' •
the initial NO in addition to the NO present initially; and NO.
£ 1
is the initial NO concentration in ppb (ppb/min)
2) T = time, in minutes, to the maximum NO0 concentration (min)
max .300 • ^
Dose = J NO2 dt where NO- is NO- ppm and t = minutes (ppm-min)
Oxidant Formation
©
Max. Rate = ' Oxi>jant where T , and T / are the times to form 3/4
3/4 ~ 1/4
and 1/4 the maximum oxidant; and max. oxidant
is the maximum oxidant concentration (ppb/min)
_\ ; „ , Max. Oxidant , m . ., ,. , . ., /0 .,
5j Avg. Rate =. r= where Tn /_ is the time to form 1/2 the maximum
I/O •'•/*'
oxidant, and Max. Oxidant is the maximum oxidant
concentration (ppb/min)
6 ) Max. Cone. = maximum oxidant concentration (ppm)
^ \s
1) T = time to the maximum oxidant concentration (min)
max
8 ) Dose =•/ Oxid. dt where Oxid. is oxidant and t = minutes (ppm-min)
^—' 0
Hydrocarbon Disappearance
Final Cone. = ppm hydrocarbon after 300 minutes irradiation (ppm)
T. __, T and T = the times required to reduce the hydrocarbon concen-
U. /o u«i) U»^o
tration to 3/4, 1/2, and 1/4 of the original (min)
HC.- HCf
(is) Max. Rate = -rrr;— ^ : where T , and T , are times for the disappearance
^3/4 ~ 1/4' '
of 3/4 and 1/4, respectively, of the hydrocarbon
disappearing in 300 minutes; HC. is the initial
hydrocarbon concentration; and HC,. is the final
hydrocarbon concentration (ppb/min)
3-2
LOCKHEED MISSILES & SPACE COMPANY. INC.
-------
LMSC-D406484
; Table 3-1 (Cont.)
HC. - HC£
fl4) Avg. Rate = —^~ where T. /0 is time for the disappearance of 1/2 the
— 1/2
hydrocarbon disappearing in 300 minutes; HC. is the
initial hydrocarbon concentration; and HC, is the final
hydrocarbon concentration (ppb/min)
,15) Max. Aid. = Maximum total aldehyde concentration (acetaldehyde for these runs)
(ppm)
Max. PAN = Maximum peroxyacetylnitrate concentration (ppm)
NO.
NO Rate = OT, l— where NO. = initial NO and T , = time to reduce NO to
1/2 1
half of original concentration (ppb/min)
NO Dose
NO0 Dose Factor = ^ Mr.— x 100 where NO = initial NO (%)
u oUU INvx X, X
X.
Ozone Dose Factor = °Z°nfJ?°Se x 100 (%)
oUU JNvJ
Xi
NO0 FWHM = Full width at half-maximum of NO0 curve (min)
2 4
Crossover Time = Time at which NO and NO2 curves cross (min)
2-21 = NO0 T - Crossover time (min)
£i IX1HX
7-21 = Ozone T - Crossover time (min)
3-3
LOCKHEED MISSILES & SPACE COMPANY. INC.
-------
LMSC-D406484
The four materials affect the behavior of the propylene/NO reaction system differently.
,A
As a summary, Figs. 3-1 to 3-4 show the photochemical results for Teflon, Pyrex,
aluminum, and stainless steel, respectively. Each figure gives the average for the two
S/V ratios and the two cleaning techniques for each spectral distribution. Run param-
eter data are given in Appendix A for each run. To account for differences in initial NO0
Lt
percentage in the NO , a linear adjustment was made to the Table 3-2 input data, as
X
discussed in Appendix A.
Table 3-2 gives the effects of the three independent variables, for the four materials
separately. Notations for that table are:
m = mean value of the parameter
A = efi'ect of changing -from low S/V value (1.3 ft ) (represented by -1) to high
S/V value (represented by +1) (2.7 ft'1)
B = effect of changing from cutoff spectrum (-) to full spectrum (+)
C = effect of changing from purge cleaning technique (-) to vacuum offgassing
(+) cleaning technique
AB, AC, BC, ABC = interaction effects
s = estimate of the standard deviation of replicates
10% = value of parameter that would be exceeded by chance 1 time in 10 if
true value of effect were zero, i.e., significant at 10% level
1% = value of parameter that would be exceeded by chance 1 time in 100 if true
value of effect were zero, i.e., significant at 1% level
The table is calculated using standard methods for factorial tests (Ref. 4). For each
material, 2 levels of A x 2 levels of B x 2 levels of C = 8 runs are available.
From these we form 4 pairs that differ only with respect to one of the independent
variables (A, B, or C). The average of the four differences is then the estimated
effect of the independent variable. This procedure is repeated three times. For this
balanced experimental plan the estimated effects are uncorrelated with each other, a
decided advantage in interpretation. Interactions are determined from the intcreffects
of the variables.
3-4
LOCKHEED MISSILES & SPACE COMPANY. INC.
-------
TEFLON
3.0
2.5
8 '-
1.0
0.5
/
FULL SPECTRUM
CUT SPECTRUM
0 50 100 150 200 250 300
TIME (MINI
Fig. 3-1 Composite Photochemical Test Fig. 3-2 Composite Photochemical Test
Results for Teflon Film Results for Pyrex Surfaces
Surfaces
ALUMINUM
100 150 200 250 300
TIME (MINI
g '•»
STAINLESS Stta
\
^^— FULL SKCnuM
---- OJTSKCnUM
100 ISO
TIME (MIN)
200 290 MO
Fig. 3-3 Composite Photochemical Test Fig. 3-4 Composite Photochemical Test
Results for Aluminum Results for Stainless Steel
Surfaces Surfaces
3-5
LOCKHEED MISSILES & SPACE COMPANY. INC.
-------
Table 3-2
EFFECTS BY MATERIAL
(1) NO2 FORMATION PARAMETERS
r
O
O
m-
m
o
t
(A
U)
r
m
U)
0)
T>
o
m
O
O
2
"£
Z
"**
Effect
• m
A
B
. C
AB
AC
BC
ABC
8
co i»%
l
Effect
m
A
B
C
AB
AC
BC
ABC
8
10%
1%
(l) N0n Rate (
Alum Pyrex
13.1 10.0
0.35 -0.7
C,
Teflon
8.63
-0.26
4.42f 5.3t 2.77
-1.38 0.5
-1.65 0.4
1.50 -1.5
-1.38 1.2
0.10 -1.2
2
3
5
-1.71
-0.09
-0.58
-0.06
0.13
.60
.87
.33
S.S.
25.4
13.5*
9.3*
-10.3*
3.1
- 1.7
- 0.6
- 1.4
Alum
106
-
-38.5*
8.0
6.0
- 7.5
4.0
4.5
@ NO» FWHM (21)
Alum Pyrex
99* 103
6.2 27*
-32.7* -59*
1.2 - 9
0.7 -23
- 5.2 -11
0.7 2
2.2 15
t,
Teflon
125
-19*
-69*
9t
17
11
2
- 3
4.83
7.20
9.90
S.S.
86
- 9T
-31*
11
- 1
1
- 5
- 1
^ r
Alum
56
- 1.7
-17.2*
5.2
4.7
- 5.7
3.7
2.7
T) NO, T
*-* 2 max
Pyrex Teflon
131 154
37.2* -19.1*
-74.2* -67.5*
- 3.2 16*
-30. 7 3
- 4.7 9
- 6.2 7
18.2 -12
7.35
11.0
15.1
Crossover Time
Pyrex Teflon
78 82
12.5* 0.7
-35.5* -26.2*
4.5 13.7
- 9.5 1.2
1.5 7.2
- 6.5 - 0.2
6.5 -4.2
4.18
6.23
8.57
(3) NO« Dose
S.S.
70
-22.5*
-28.0*
20.5*
1.0
- 0.5
- 8.0
1.0
S.S.
38
-14.0*
-13.5*
13.0*
2.0
- 1.5
- 4.0
1.5
^-^ 6
Alum Pyrex
176 178
14T 85*
-32* -76*
10 -24*
-10 -41
0.2 -29
- 5 21
1 33
8.43
12.6
17.3
© = ©
x^x v>>
Alum Pyrex
50 58
1.7 26.0*
-21.2* -38.5*
2.2 - 7.5*
1.2 -20.5
- 1.7 - 6,0
0.2 1.5
1.7 13.5
2.91
4.33
5.97
Teflon S.S.
219 153
-47* - 4
-70* -41*
- 2 14t
14 - 4
11 4
15 - 6
-7 - 1
(fV
v_3'
Teflon S.S.
70 36
-25.7* - 8*
-35. 7* -14*
7.7* 8
6.7 - 1
6.2 1
1.2 - 4
-14.2 - 1
@ NO. Dose Factor
Alum
39.0
2.5
-7.7*
1.7
-2.5
-0.6
-0.8
0.5
' Z
Pyrex
39.2
16.7*
-10.3*
- 6.0*
-10.5
- 5.0
2.7
6.0
1.
2.
3.
Teflon
48.7
-10.5*
-16. 1*
- 1.8
4.2
2.1
3.3
- 1.5
93
88
96
S.S.
34.5
-0.7
-9.0*
4.2*
-1.2
0.2
-2.7
-0.4
(17) NO Rate
Alum
8.71
-0.08
2.60f
-0. 13
-0.24
0.78
-0.23
0.09
^-^
Pyrex
8.55
- 2.06f
4.61*
1.23
0.38
- 0.92
1.56
- 1.29
1.
1.
2.
Teflon
7.20
0.59
2.45f
- 0.16
- 0.01
- 0.30
- 0.06
0.33
29
9*2
69
S.S.
14.4
3.8*
4.8*
-1.9T
0.4
-0.3
0.7
-0.2
f
O
fe.
4-^
O
rrx
*Signifleant at 1% Level.
tSignificant at 10% Level.
-------
Effect
Ozone Max Rate
Table 3-2 (Cont.)
4
(2) OZONE PARAMETERS
(5) Ozone Avg. Rate © Max Ozone Cone
Ozone T
Max
r
O
O
7\
I
m
m
o
2
(D
en
p-
m
U)
B>
en
J
O
m
O
O
2
J>
Z
"*
m
A
B
C
AB
AC
BC
ABC
8
10%
1%
co
-lj Effect
m
A
B
C
AB
AC
BC
ABC
8
10%
1%
Alum
24.7
-9.7*
7.7*
0.9
-3.7
-2.0
-1.1
0.9
-
Pyrex Teflon
24.7 21.3
-8.5* -1.4
8.2* 11.2*
4.5t 2.4
4.4 -5.6
1.0 -1.2
-0.4 -2.4
-1.0 -0.5
2.66
3.97
5.45
S.S.
21.2
-8.2*
10.8*
0.4
-1.9
-0.1
-0.5
1.1
(16) Max Pan Cone
Alum
0.35
-0.05
-0.11*
0.03
0.09
-0.02
0.04
-0.06
*Significant at 1%
Pyrex Teflon
0.14 0.29
0.04 -0.10*
0.02 0.19*
-0.02 0.12*
0.03 -0.12
-0.06
0.03 0.03
-0.01 -0.02
0.039
0.058
0.080
Level.
S.S.
0.23
-0.03
0.01
0.02
0.07
0.01
0.03
-
Alum Pyrex
4.08 3.60
Teflon
3.12
- 0.66* - 1.35* 0.20
1.61* 2.18* 1.92*
- 0.08 0.56* 0.16
- 0.21 0.01
0.25 - 0.05
- 0.13 0.39
- 0.56
© =
Alum Pyrex
116 115
8.5T 46.0*
-39.0* -62.0*
2.5 - 6.5
- 3.0 -34.0
- 3.5 -16.5
1.0 5.5
1.0 20.5
4
7
9
- 0.11
- 0.21
- 0.20
0.08
0.258
0.385
0.529
(7) - (2
Teflon
148
-30.5*
-62.5*
- 1.0
14.5
- 0.5
0.5
- 4.5
.86
.24
.96
S.S.
5.68
- 0.01
3.35*
- 1.05*
- 0.01
- 0.05
0.02
0.01
Alum Pyrex
1.00 0.93
- 0.14* - 0.12*
0.04 0.09*
0.08* 0.13*
0.02 - 0.01
0.03
0.01
- 0.03
0
0
0
Teflon S.S.
1.03 0.90
-0.06 - 0.26
0.20* 0.40*
0.15* O.OSt
-0.06f - 0.03
-0.04 - 0.02
-0.05 0.01
0.02
.033
.049
.068
l) (8) Ozone Dose
S.S.
103
- 4
-31*
11*
-
- 3
- 6
2
Alum Pyrex
156 126
-22.0* -48*
38.0* 61*
5.5 26*
-11.0 15
10.5 9
- 8.5 3
6.5 -13
8.
12.
17.
Teflon S.S.
131 157
4.0 -29*
94.5* 48*
9.5 - 4
21.0 -12
14.5 2
11.5 8
8.0 3
51
7
4
Alum
172
6
-56*
8
2
- 9
4
3
©
Alum
34.3
- 5.4*
7.9*
0.9
- 2.7
1.8
- 1.7
1.5
Pyrex Teflon
189 230
-58* -31*
-98* -89*
- 2 12f
-43 15
-15 7
- 1 1
27 8
8.05
12.0
16.5
S.S.
137
-18*
-45*
25
1.2
- 4
-10
4
Ozone Dose Factor
Pyrex Teflon
27.8 28.6
-12.5* 1.9
12.3* 20.6*
4.2* 0.8
2.0 - 3.6
2.9 - 3.4
- 0.7 - 2.5
- 3.1 1.7
1.68
2.50
3.44
S.S.
34.1
- 6.4*
10.8*
- 3. Of
-
-
0.8
0.6
t Significant at 10% Level.
r
en
O
o
oo
-------
Table 3-2 (Cont.)
(3) PROPYLENE PARAMETERS
o
o
X
I
m
m
D
en
r
m
(A
O
m
O
O
2
TJ
co
oo
Effect
m
A
B
C
AB
AC
BC
ABC
s
10%
1%
Effect
m
A
B
C
AB
AC
BC
ABC
8
10%
1%
(9) HC Final Cone.
Alum
0.16
0.05
-o.oet
-
-0.01
-0.01
0.04
0.03
Pyrex
0.26
0.15*
- 0.19*
- 0.10*
- 0.06
- 0.10
0.04
0.06
0.
0.
0.
Teflon
0.26
-0.10*
-0.24*
-0.02
0.04
0.04
0.08
-0.08
039
058
080
S.S.
0.15
-0.03
-0. 09*
-
0.04
-0.04
-0.02
0.02
Alum
94
- 3
-28*
9
9
- 3
9
- 8
(O) HC Max. Rate
Alum
21.8
-3.7t
6.6*
-0.8
0.8
2.9
1.2
-0.8
Pyrex
23.6
-10.2*
13.5*
0.9
- 0.5
0.8
- 0.2
2.0
Teflon
15.9
4.0*
8.2*
2.2
1.1
-1.0
0.3
0.6
1.94
2.89
3.98
S.S.
23.9
0.5
9.2*
1.4
-0.4
0.1
1.3
-0.2
Alum
11.5
-0.85
4.04*
-1.52
-0.88
1.3
-1.06
0.53
(lO) HCT0.75
Pyrex
117
27*
-56*
- 3
-18
-11
4
18
8
12
17
@ HC
Pyrex
10.1
- 2.2*
- 5.3*
1.1
1.0
-0.5
- 1.4
-
1
1
2
Teflon
124
-
-38*
11
- 4
-
- 1
-
.34
.4
.1
Av. Rate
Teflon
8.2
0.6
3.6*
- 0.3
0.1
0.1
- 0.5
0.5
.06
.58
.17
S.S.
64
-21*
-20*
22*
4
-11
- 9
7
S.S.
16.2
4.5*
5.8*
- 3.0*
0.3
-
0.6
- 0.3
Alum
133
3
-43*
12 1
7
- 7
7
- 2
Alum
0.87
- 0.13
- 0.08
- 0.10
- 0.06
- 0.04
0.04
- 0.06
Qj) HC T 0. 5 (lj) HC T 0. 25
Pyrex
157
44*
-76*
- 8
-25
- 9
- 1
22
6.85
10.2
14.0
@ Max.
Pyrex
0.70
- 0.40*
0.11
-
- 0.09
- 0.01
0.07
- 0.03
0.
0.
0.
Teflon S.S. Alum Pyrex Teflon S.S.
180 97 166 189 218 133
-15* -22* 8 45* -19* -18*
-70* -34* -48* -80* -84* -45*
6 19* 7 -15* 4 19*
37 6 -26 6 4
7 - 7 -10 -14 7 -11
4-9 7 5 7-11
-8 3 2 26 -9 5
6.76
10.1
13.9
Aid
Teflon S.S.
0.75 0.76
- 0.01 - 0.02
0.06 0.01
- 0.02 - 0.06
- 0.07
- 0.06 0.04
- 0.02 0.02
- 0.05 - 0.10
135
201
277
*Signlficant at 1% Level.
tSignificant at 10% Level.
en
O
d
*>•
o
>£>•
00
-------
LMSC-D406484
For example, consider the following data for NO0 time-to-maximum for Teflon
A
surfaces:
Run No. ABC Time
38 + + + 121
34, 35 + + - 103
42 - + + 142
43, 44 - + - 115
36, 39 + - + 192
37 + 162
40 - - + 193
41 - - - 205
Effect A is [(121-142) + (103-115) + (192-193) + (162-205)]/4 = -19.1
Effect B is [(121-192) + (103-162) + (142-193) + (115-205)1/4 = -67.5, etc.
Note the efficiency with which each data point is utilized. This is characteristic of full
factorial experimental designs. Selected conditions were replicated, to provide data on
reproducibility to determine whether effects are "real" or are due to random deviation
in chamber behavior. In this analysis, these replicates are averaged to
allow the orthogonal data treatment just described.
Deviations between replicates were used to obtain s, the estimated standard deviation
of the parameter. For all four materials and 23 parameters, pooling of the deviations
is justified. The significance levels are then calculated from the t value for the num-
ber of degrees of freedom and s. The "effects" found are compared to the 1-percent
and 10-percent significance levels in data interpretation.
Practical considerations required that the experimental plan be conducted for each mate-
rial and each S/V level as a subgroup. A time trend analysis (Appendix B, page B-10)
was thus conducted, to see if systematic drift was present in the experiment. Three sets
of replicates are available, immediate reruns, reruns differing by more than 1 and
less than 15 run numbers, and a complete rerun of the first material used in the experi-
ment (run number differing by more than 50). Both the Rank Sum Test and the Sign
Test show no evidence of drift, either within each material tested, or on the basis of
material block-to-material block.
3-9
LOCKHEED MISSILES & SPACE COMPANY. INC.
-------
LMSC-D406484
3.1 DETAILED STATISTICAL ANALYSIS
A detailed statistical analysis of the raw data was performed. The ground rule was that
all photochemical runs submitted for analysis were used. The experimenter was free
to declare a run invalid on the basis of physical grounds, such as equipment failure, un-
due deviation from desired initial conditions, suspected off-calibration instrumentation,
or new material subject to aging effects. The analysis results are given in Appendix B
in some detail and are summarized here.
3.1.1 Covariate Analysis
The covariate analysis considered the following initial conditions: %NO2> HCinit,
NOV, HC/NOV, NO0 and T ,. as covariates, where:
«A. .A. ^
%NO2 = percent of NO2 in initial NOX
HC Init = initial propylene concentration, ppm
NC»X = initial NOX concentration, ppm
HC/NOX = ratio of propylene/NOx initially
NO2 - initial concentration of NO2, ppm
correction factor to NO2 time-to-]
in %NO? from desired value of 10%
T ,. = correction factor to NO time-to-maximum to account for deviation
As a single covariate correction, %NO0 or T ,., were very similar in effectiveness in
2 ^
increasing R (Appendix B, page B-23). As the %NO2 correction is based purely on
statistical grounds, and as T ,. was based purely on observation of the photochemical
behavior, this is gratifying. The result further emphasizes the importance of con-
sidering initial %NO0 when correlating smog chamber results. For most of the mate-
2
rials and the run parameters, the increase in R obtained by including covariate
correction terms beyond %NO2 and HC Init was insignificant. Therefore the bulk of the
statistical analysis was conducted on the raw data as corrected by these two covariate
correction factors.
3.1.2 Principal Components
The 23 parameters used to describe each photochemical run are very strongly
interrelated (See Tables B5-2 through B5-4 of Appendix B.) In particular a
3-10
LOCKHEED MISSILES & SPACE COMPANY. INC.
-------
LMSC-D406484
single "TIME" value can be used to represent the following: 1/NCL Rate, NO2 Time
to Max, I/Ozone Avg. Rate, Ozone Time to Max, HC Time 0.5, 1/HC Avg. Rate,
I/NO Rate, NO0 FWHM, and Crossover Time. Time to 75% HC disappearance cor-
£i
relates very well with 50% disappearance time (about .98) and time to 25% disappear-
ance correlates almost as well (.95). These are the parameters that form the principal
component called TIME, as defined in Table B5-5. It was also found that a combination
of NO2 and Ozone Doses and Dose Factors can be used, rather than each separately.
These are the principal component called DOSE. The maximum rates and concentra-
tions form another pair of principal components. These two combine HC Max. Rate,
HC Final Cone., Ozone Max. Rate, and Ozone Max. Cone, by two different linear
combinations, and are called MISC I and MISC II.
These four principal components were studied as run results themselves. In particular,
cross plots of the four, taken two at a time, were prepared and the effects of S/V,
Spectrum, and Cleaning noted. The patterns obtained were quite regular.
3.1.3 Effects by Materials
Models were formulated for each of the 23 parameters for each material separately, as
well as for the material-to-material contrast. These models differ from the orthogonal-
ized analysis just reported in using all the replicates as separate data points, not aver-
aging. An effects-by-materials table was then prepared, and is given in Table 3-3.
This table was calculated from results of the analysis of variance within each material
using the Doolittle Matrix Inversion Method. For Table 3-3, s is the square root of the
unexplained variability (experimental error), which differs significantly for each
material. Table 3-3 differs from Table 3-2 in (1) using all replicates as separate data
points, not averaging; and (2) using the covariate adjusted data.
Effects are declared significant at the 10% or 1% level using similar criteria
to those used for Table 3-2. The results of the two analysis methods are
very similar, which further reinforces the statement that the data is of
sufficiently high quality and validity as to support models which can be used
for reconcilliation of data from other smog chambers, and as the basis for
3-11
LOCKHEED MISSILES & SPACE COMPANY. INC.
-------
TABLE 3-3
Effects by Material
5
0
I
m
m
0
Z
(A
(A
m
(A
(A
T)
0
m
0
O
Z
TJ
Z
•<
uo
fe
Effect
m
A
B
C
AB
AC
BC
ABC
s
10$
1%
Effect
m
A
B
C
AB
AC
BC
ABC
s
10$
1$
1 1/HOgRate
Alum
.077
-.007
-.026t
.009
.012
-.007
.005
.008
.016
.025
.061
Alum
. 39.1
. 2.90t
-7.67*
1.84
-2.47t
-.596
-.529
.178
1.38
2.12
5-27
Pyrex
.107
.012t
-.058t
.004
-.008t
.Ol4t
-.013t
.006
.0016
.007
.070
18 N02I
Pyrex
39-3
16,6*
-18.2*
-5-85*
-10.3*
—5-25*
.2.45*
5-95*
. 0
.000007
.000007
Teflon
.114
.005
-.042*
.025*
0.0
.012t
-.002
-.005
.0049
.007
.018
tose Factor
Teflon
48.8
-10. 5t
-16.2*
-1.68
.4.40
:.=2.20.
3.45
-1.58
3.46
5-19
12.9
S.S.
.043
-.021*
-.018*
.021*
.002
-.005t
-.007t
.003
.0043
.005
.009
S.S.
34.7
•-.600
-8.95*
4.25*
-1.35
.150
-2.70t
- .300
1-30
1.46
2.79
2
Alum
100.9
.833
-36. 8t
9.17
7-17
-4.83
3.83
4.83
12.5
19.2
47.7
20
Alum
99.8
6.04
-33-3+
.960
.208
5-54
• 791
1.29
9.36
14.4
35-7
2 max
Pyrex
126.7
4l.lt
-71.lt
-.624
-25 . It
-5-87
-6.12
20. 6t
3-54
15-3
154.3
H02 FWHM
Pyrex
102.8
26. 9t
-59- 4t
-9-87t
-22 . It
-11. 6t
1.13
15 .4t
2.12
9.16
92.4
Teflon
147.2
-20. 6t
-67.4
16. 9t
2.63
10.9
9.13
-11. 9t
7-92
11.9
29.5
Teflon
124.8
-19-1*
-69.6*
-8.88t
17.6*
10.87t
2.88
-3-38
3.90
5-85
14.5
S.S.
66.2
-18.5*
-26.1*
19.4*
-.875
.625
-6.12t
.625
4.86
5-46
11-7
S.S.
86.4
-9.04*
-31-5*
11.8*
-.800
.542
-4.96t
-.708
2.49
2.80
5.34
3
Alum
177-5
15.0
-33. 8t
9-46
-11.8
-1.54
-3-71
-2.71
10.4
16.0
39-7
21
Alum
51-1
-1.29
-15- 5t
5.79
6.04
-3-71
3.46
2.96
8.35
12.8
31-9
N02 Dose
Pyrex
178.0
82.0*
-74-5*
-24.5*
-42.5*
-31.5*
16.5*
32.0*
0
.00003
.0003
Crossover
Pyrex
68.9
13-2
-33- 7t
5-25
7-25t
.750
-7.25t
7-25t
l.4i
6.10
61.5
Teflon
220.1
-46.3*
-72.5*
-1.50
16. ot
11.0
18.7
-7.75
7-37
11.1
27.4
Time
Teflon
75-2
.625
26.1*
13-1*
1-13
8.88*
1.62
-3-63t
2.20
3-30
8.19
S.S.
155.5
-5.08
-40.6*
14.6*
-3.58
2.58
-6.92
.083
6.64
7-45
14.2
S.S.
29.5
-12.1*
-12 . 1*
11.6*
• 917 p5
-.417 £
-3.43 |
• 583 £
2.54
2.85
5-44
-------
TABLE 3-3
Effects by Material (Continued)
Effects
0
O
T
CHEED Ml
CO
CO
r
m
CO
CO
TJ
O
m
COMPANY
uo
i
u>
m
A
B
C
AB
AC
BC
ABC
s
10$
1$
Effects
m
A
B
C
AB
AC
BC
ABC
s
10$
23
Alum
116.6
8.l7t
-38.7*
2.83
-3.67
-3-17
.667
.667
3-30
5.08
12,6
9
Alum
•159
.048
-.073
-.008
-.013
-.018
- .043
.023
.050 .
.077
• 191
=7-21 8
Pyrex
114.9
46. 6t
-62.lt
-5-62
-31.4t
-16. 9t
4.38
21.lt
2.12
9.16
92.4
HC Final
Pyrex
.254
.l47t
-.198t
-.127t
-.O87t
-.098t
• .047
.o67t
.014
.061
.610
Teflon
147.1
-31-6*
-62.9*
- .125
14.3*
-.375
.875T
-4.38*
.407
.610
1-51
Cone
Teflon
.269
-.092t
-.245* ;
-.025
.045
.025
.O85t
-.O88t
.025
.037
• 093
S.S.
103-3
-2.67
-30.9*
li.lTt
-1.17
-2.58
-5-23
1.92
5.94
6.67
12.7
S.S.
.151
-.039
-.O89t
.006
.052
-.044
-.024
.026
.052
.058
.111
Alum
151.8
-19 -9t
35- 2t
2.20
-14.6
5.45
-9-35
-4.60
10.6
16.3
40.5
10
Alum
88.4
-3.17
-26. 8t
9-17
9-17
-1.83
8.83
-8.17
9-10
14.0
34.7
Ozone Dose
Pyrex
128.8
-52.lt
62. 9&
24. 2t
17- 7t
3-52
. 4.42
-12.5
3-68
15-9
l6o.4
HC T 0
Pyrex
110.0
29.0
-55 -Ot
-4.0
-16.0
-10.0
2.0
20.0
9-90
42.8
431.5
Teflon
129.6
7-36t
88.7*
10. 9t
-18.2*
-12 .4t
-4.09
7-83t
4.19
6.28
15-6
• 75
Teflon
118.2
1.38
-38. 9t
10.9
-1.87
.875
1.62
-.218
10.5
15.7
39-1
S.S.
160.7
-28.3*
48.2*
-3-54
-11. 9t
.858
S.Olt
4.8i
6.63
7.45
14.2
S.S.
60.4
-18.7*
-20.4*
20.2*
4.04
-9.29t
-8.54t
4.96
5.61
6.30
12.0
Alum
33-3
-4.89t
7.54*
.192
-3.06t
.892
-1.84
-.342
1.66
2.55
6.34
Alum
128.4
4.35
-40.2*
12.2
8.75
-4.75
6.25
-1-75
9-96
15.3
38.0
19 Ozone
Pyrex
28.1
-13-lt
12. 6t
3-66
1.89
2.13
-.087
-2.8l
• 919
3-97
4o.l
11 HCT 0.
Pyrex
150-9
47. 6t
75. 6t
-6.62
-18.9
-8.87
-2.12
24.6
9.19
39-7
400.5
Dose Factor
Teflon
28.7
2.01
20.0*
1.21
-3-54t
-3-36t
-1.84
1.69
1.44
2,16
5.36
5
Teflon
171-9
-15.lt
-70 . 6*
7.62
3.38
8.62
6.12
-7-88
5.79
8.68
21.6
S.S.
35-7
-6.60*
10.8*
-1.05
-2.96t
-1.72
-.421
-.445
1.98
2.22
4.24
S.S.
93-2
-18.4*
34.6*
17-6*
4.62
-3-62
-7-38t
.138
4.11
5-74
ll.O
-------
TABLE 3-3
Effects by Material (Continued)
5
O
I
m
m
D
2
0)
to
r
m
to
R>
to
TJ
0
m
0
O
2
T)
Z
Effect
m
A .
B
C
AB
AC
BC
ABC
s
1036
&
Effect
m
A
B
C
AB
AC
BC
ABC
s
lOjfc
&
Alum
.114
.oo4
-.032t
.002
.010
-.007
.001
.006
.0089
.014
.034
17 I/NO
Pyrex
.127
.o4ot
- .o69t
-.007
-.025t
.004
-.009
.014
.0037
.016
.161
Rate
Teflon
• 139
-.013
-.053*
.003
.003
.011
0.0
-.009
.0103
.015
.038
4 1/0 zone
S.S.
.069
-.016*
-.026*
.018*
.004
0.0
-.007t
0.0
.0048
.005
.010
Alum
.044
.Ol7t
-.012
.002
.002
.004
.002
0.0
.0105
.016
.040
6 Max Ozone Cone.
Alum
1.01
-.139*
.044t
.0?9t
.001
-.001
.004
.014
. .021
.032
.080
Pyrex
..943
-.125t
.110
.150t
.010
.010
-.005
-.035-
.028
.121
1.22
Teflon
1.02
-.061
.197*
.159*
-.05lt
-.034t
-.04 it
.006
.019
.028
.071
S.S.
• 921
-.246*
.207*
.048*
-.046*
-.014
.013
.036t
.02"!
.024
.045
Alum
167.3
7.08
-54.1*
9.42
3.42
-6.08
4.08
3-60
11.1
17.1
42.4
Pyrex
.046
.023t
-.022t
-.012t
-.Ol4t
-.008
.008
.008
.0021
.009
.092
Max Rate
Teflon
.051
-.006t
-.028*
-.Ollt
.Ol4t
.007t
.out
-.003
.0033
.005
.012
5 I/Ozone
S.S.
.052
.022*
-.027*
-.001
-.006t
.002
.001
-.003
.0034
.004
.007
Alum
• 251
.028
-.098*
-.011
-.003
-.018
.002
• .019
.021
.032
.080
7 Ozone Tmax
Pyrex
183.6
6l. 2t
-95. 3t
-.250
-38. 2t
-16.2
-2.75
28. 3t
4.24 .
18.3
184.8
Teflon
221.9
-31. 4t
-88.8*
13 At
15- 9t
8.62
2.62
-8.13
8-75
13.1
32.6
S.S.
132.4
-l4.9t
-43-7*
23.2*
-.458
-2.04
-8.29t
3A6t
7.06
7-93
15.1
Alum
.365
-.050
- . 122t
.030
.078t
-.030
.032
-.o88t
.037
.057
.141
Pyrex
.313
• 155t.
-.217t
-.067t
- . 108t
-.o47t
.024
• O73t
.007
.030
• 305
l6 Max Pan
Pyrex
.143
.039
.026
-.014
-.001
-.021
-.oo4
-.011
.0495
.214
2.16
Avg. Rate
Teflon
-350
-•039t
-.228*
-.048*
.038t
.020t
.048*
-.037t
.012
.018
.045
Cone
Teflon
.291
-.098t
.172*
.138t
-.108t
-.053
.047
-.023
.038
• 057
.141
S.S.
.182
-.004
- . 117*
.041*
.003
.003
-.026t
-.004
.013
.015
.028
S.S.
.251.
-.027
.013
.022 K
.058t ?
.002 If
-.018 §
.007
.032
.036
.069
-------
TABLE 3-3
Effects by Material (Continued)
LOCK
HEED
2
p
z
J
J
1
Effect
m
A
B
C
AB
AC
BC
ABC
s
1036
1$
Effect
m
A
B
C
AB
AC
BC
ABC
s
10$
1$
Alum
160.8
8.32
-46.0*
8.37
7.13
-7.63
6.88
2.88
10.1
15'. 5
38.6
Alum
.880
-.126t
-.069
.001
-.064
-.024
.039
-.056
" .066
.102
. .252
12 HC T 0
Pyrex
185-3
54. Ot
-84. 5t
-9-50
-26.0
-8.00
-1.50
23.0
9-90
42.8
431.5
•25
Teflon
210.9
-19 At
-83.6*
5-12
6.12
8.37
8.62t
-9-13t
5.6l
8.51
20.9
15 Max. Aid
Pyrex Teflon
• 719
-.027
.142
.047
-.023
-.048
.062
.-.125
.042
.182
1.831
• 773
-.034
.064
.001
-.006
-.029
-.021
-.o4i
.091
.136
• 339
S.S.
129.8
-14.5*
-45.3*
17.3*
2.46
-7.46t
-9-71t
3-54
5-31
5.96
11.4
S.S.
• 77^
.020
.015
-.068
-.097
.060
.044
-.103
: .106
• 119
.227
13
Alum
.048
.009
-.015t
.002
-.004
-.006
-.002
.004
.007
.010
.027
Alum
49.9
1-75
-21.3*
3-25
• 750
-1-75
.250
1.25
4.32
6.64
16.5
1/HC Max Rate
Pyrex
.050
.026*
-.031*
0.0
-.008*
-.002t
0.0
-.003t
.0001
.0004
.004
22 = 2 -
Pyrex
57.8
26. 3t
-38.3*
-6.37t
-18. 6t
-6.63t
.625
13 At
• 707
3.06
30.8
Teflon
.069
-.021t
-.036*
-.012
.005
.012
.005
-.006
.009
.013
.003
21
Teflon
69.5
-23-3*
-36.0*
8.50*
6.75*
6-75*
l.50t
-14.3*
•577
.865
2.15
S.S.
.044
.001
-.017*
-.002
0.0
.001
0.0
0.0
.003
. .003
.006
S.S.
36.6
-7.54*
-14 . 3*
8.04*
-1.54
.792
-3A6t
-.204
2.46
2.76
5.72
14 l/HC Avg. Rate
Alum
.090
.001
-.024t
.007
.004
-.007.
.001
.002
.010
.015
.038
Pyrex
.106
• 037t
-.059t
-.011
-.020
-.008
.002
.019
.006
.026
.262
Teflon
.124
-.Ol4t
-.058*
.002
.003
.005
.008t
-.008t
.005
.007
.019
S.S.
.064
-.Ol5t
-.025*
.014*
.005t
-.oo4t
-.007*
.002
.003
.003
.007
-------
LMSC-D406484
further eluciation of the photochemical reaction mechanism.
3.2 MATERIAL DIFFERENCES
The four materials may be grouped in terms of increasing NO0 formation rate: Teflon,
Z
Pyrex, aluminum, stainless steel. Pyrex and aluminum are similar in behavior for
most parameters. Other reactivity manifestations such as times to NO0 maximum, 50
£
percent propylene destruction, and NO,, dose follow the same order. Figures 3-1 to 3-4
provide a better understanding of material differences than word description, but also
see the means by material in Tables 3-2 and 3-3.
3.3 EFFECTS OF FACTORS
Of the three independent variables studied, the spectral change (Effect B) caused the
largest change in behavior. Table 3-4 gives the significant effects count for the four
materials and the three variables. Spectral distribution was significant at the 1% level
(one chance in 100 that it was really not significant) for nearly all 23 parameters for all
four materials. The notable exception was NO2 formation rate, in which case spectrum
was significant at the 1% level only for stainless steel, and at the 10% level for aluminum
and Pyrex, and not significant for Teflon. Recall that all runs were performed at con-
stant k,. Maximum ozone concentration was strongly affected for the Teflon and stain-
less steel, which are the two materials with highest ozone levels, as shown in Table 3-5
below. Spectrum is significant at the 1% level for all but aluminum.
For all four materials, the cutoff spectrum consistently and clearly slowed the reaction
relative to the full spectrum. Table 3-6 shows the ratio of cut to full spectrum for
several "reactivity" measures.
S/V ratio (Effect A) was next in importance in affecting smog chamber behavior. Pyrex
results were influenced by S/V to a larger degree than were the other materials.
Cleaning had its largest effect for stainless steel, and Pyrex. These counts of signifi-
cant effect are somewhat misleading, in that some of the calculated run parameters,
such as NOg Time Max. and Ozone Max. Cone, are of greater practical importance
than others. As just shown in Table 3-5, Teflon S/V hardly affects Ozone Max Cone.,
and the general picture is that S/V is of less importance for Teflon than for, say, alumi-
num. To make such judgments, the run graphs, which are given in Appendix A, should
be consulted.
3-16
LOCKHEED MISSILES & SPACE COMPANY, INC.
-------
LMSC-D406484
Table 3-4
SIGNIFICANT EFFECTS BY MATERIAL
S/V
Significant at
Spectrum
Significant at
Cleaning
Significant at
Material
Aluminum
Pyrex
Teflon
Stainless
Steel
1%
5
20
12
14
10%
3
1
-
1
1%
18
20
20
21
10%
3
1
1
-
1%
1
9
4
13
10%
1 -
2
2
4
Table 3-5
MAXIMUM OZONE CONCENTRATIONS (PPM)
Aluminum
Pyrex
Teflon
Stainless Steel
Averages
All runs
Full Spectrum
Cut Spectrum
High S/V
Low S/V
Vacuum Clean
Purge Clean
1.00
1. 02
.98
.93
1.07
1. 04
.96
.93
.98
.89
. 87
.99
1.00
.87
1. 03
1. 13
.93
1. 00
1.06
1. 11
.96
.90
1. 10
. 70
. 77
1.03
.92
. 88
N00 T 1
2 max
Ozone T 7
max
Table 3-6
CUT/FULL SPECTRUM RATIO
50% Propylene Destruction 11
Teflon Pyrex Aluminum Stainless Steel
1.58 1.78 1.54
1.49 1.68 1.47
1.50 1.64 1.47
1.49
1.39
1. 43
3-17
LOCKHEED MISSILES & SPACE COMPANY. INC.
-------
LMSC-D406484
3.4 OZONE DECAY RESULTS
One parameter frequently used as a measure of chamber cleanliness is the ozone half-
life in the smog chamber. Measurements were made of the ozone half-life in the cham-
ber each time a new set of surface materials was installed, and when the S/V was
changed. The conditions were: temperature, 95°F; relative humidity, 55°F dewpoint;
total HC as methane, <0.1 ppm. The initial ozone concentration was 1 to 2 ppm.
These tests were usually conducted in conjunction with the vacuum chamber off-gassing
that was used as the final cleaning step after installing the new material. Ozone decay
was determined in the dark and for the illuminator (full spectrum configuration) at its
nominal 6,500-W power output (decay in the light). Results are shown in Table 3-7.
Ozone decayed fastest in the presence of stainless steel surfaces and most slowly in
the presence of Pyrex. It should be recognized that ozone decay behavior is a function
of previous conditioning as well as material and configuration, and, by itself, has been
the subject of several research investigations such as Sabersky (Ref. 19).
Table 3-7
OZONE HALF-LIFE STUDY
Configuration
Base Chamber
Aluminum High S/V
Aluminum Low S/V
Pyrex High S/V
Pyrex Low S/V
Teflon High S/V
Teflon Low S/V
Stainless Steel High S/V
Stainless Steel Low S/V
Half-Life in
Dark (min)
430
270
340
360
340
295
350
160
190
Half-Life in
Light (min)
180
210
215
300
275
200
270
100
120
3-18
LOCKHEED MISSILES & SPACE COMPANY. INC.
-------
LMSC-D406484
The correlation coefficient for the ozone half-life and the calculated run parameters
is given in Table 3-8. The full spectrum data averaged over both cleaning methods
was used, which gives a sample size of eight. Several of the run parameters
show a significant correlation at the 5% level (1 chance in 20 that it really is
not significant) and several more at the 10% significance level. The correlation
coefficient is usually higher for ozone half-life in the dark than in the light.
This result implies that it may in some way be possible to correlate smog
chamber behavior using ozone decay in the dark as the link between chambers.
However note the poor correlation for Ozone Max. Cone, and ozone half-life.
3. 5 RELATIVE HUMIDITY EFFECTS
A numbeir of tests were conducted at lower relative humidity than the standard
condition. These were done for the purpose of obtaining additional insight
into the mechanisms that might be operative to cause the spectral effects and
to obtain experimental data on this effect for which opposite results have been
obtained by different groups (Ref. 15). The resulting data is briefly summarized
in Table 3-9. Appendix A gives the run graphs and the complete run parameter
calculation results. Decreasing the dew point slowed the photochemical re-
action for the stainless steel and aluminum surfaces. This is in accordance
with Altshuler and Bufalini hypothesis (Reference 15) that lower dew point
allows more active wall sites, which lowers the free radical concentration
in the gas phase, leading to decreased reactivity. Maximum ozone concentration
increased for all three materials studied.
For Teflon surfaces, which gave the slowest reacting system studied, a
decrease in dew point increased the reaction rate. The three data points
taken do not show a monotonic trend. Whether this is caused by experimental
error or by some trend turnaround between 17 and -20 F dew point has not
been established. The latter hypothesis is favored by the author.
For all materials and all spectral conditions, the steady state ozone
concentration is higher as the dew point is decreased. The humidity effect
occurs for both spectral conditions. This implies that the spectral effect
is np_t due to O( D).
3-19
LOCKHEED MISSILES & SPACE COMPANY. INC.
-------
LMSC-D406484
Table 3-8
CORRELATION OF OZONE HALF-LIFE WITH RUN
PARAMETERS AVERAGED OVER CLEANING
Variable
HCT 50
O3TM
FWHM
XTIME
1/Oz Avgr
1/HC Avgr
1/NO2R
I/NO RATE
NO2 DOSE
NO2 DF
O3 DOSE
°3DF
1/O3 MAX R
1/HC MAX R
O3 MAX C
HCFC
PAN Max
HCT 75
HCT 25
NO2XT
°3XT
ALD Max
TIME
DOSE
Max 1
Max 2
Sample Size
Significant Correlation (5%)
(10%)
Correlation of
Half -Life in Light
.775
. 739
.579
.346
. 806
.797
. 726
.795
.741
.411
. 307
-. 373
-.423
-. 270
.296
. 122
. 703
-.078
.767
.712
.629
.195
. 119
. 730
-.414
.009
. 349
8
. 707
.621
Correlation of
Half -Life in Dark
. 792
. 750
.629
.436
. 825
. 812
.767
. 824
. 780
.480
.399
-. 226
-.275
-.417
. 300
. 256
.539
.002
.793
.718
.655
.275
.343
.759
-.403
. 141
.437
8
. 707
.621
3-20
LOCKHEED MISSILES & SPACE COMPANY, INC.
-------
LMSC-D406484
Table 3-9
RELATIVE HUMIDITY VARIATION EFFECTS
Dewpoint NO9 Time HC
Ozone
Max Cone. Run
Material
Alum
Alum
Alum
Alum
Alum
Alum
Stainless
Stainless
Stainless
Stainless
Stainless
Stainless
Stainless
Teflon .
Teflon
Teflon
Teflon
Teflon
S/V
Low
Low
High
High
High
High
Low
Low
Low
Low
Low
High
High
Low
Low
Low
Low
Low
Spectrum
Full
Full
Full
Full
Cut
Cut
Full
Full
Full
Cut
Cut
Full
Full
Full
Full
Full
Cut
Cut
(UF)
55
22
55
19
55
14
55
22
-20
55
-12
55
12
55
17
-20
55
15
to Max Min
76
86
86
105
117
155
60
60
80
81
115
39
57
118
67
77
205
190
T. 50
92
113
113
126
145
182
81
77
105
112
135
69
83
142
93
106
230
210
(ppm)
1. 04
1.16
.94
1.00
.84
.94
1. 14
1. 14
1.23
. 87
1. 06
. 83
.90
1.08
1.20
1. 31
.80
1.03
No.
75
76
69 & 85
89
82 & 84
88
61 & 66
64
65
60 & 67
68
51, 52,
58
59
46
47
49
41
48
3-21
LOCKHEED MISSILES & SPACE COMPANY. INC.
-------
LMSC-D406484
3. 6 BUTANE RUNS
A series of tests were conducted using the n-butane/NO system. The major
X.
objectives of these tests were to determine if the spectral effect and S/V
effect were detectable in this low reactivity paraffinic hydrocarbon system.
It was felt this would help to establish the generality of the effects. The
butane-NOx system has been investigated by Altshuler (Ref. 16). Appreciable
oxidant production occurs at a restricted range of values of the HC/NOX ratio
of about 5-10 (molar basis). The oxidant production vs HC/NOX ratio curves
for butane show sharp peaks, in contrast to similar curves for propylene as
the hydrocarbon. The condition chosen for this study was 3 ppm butane and
0. 6 ppm NOX, for a HC/NOX ratio of 5 (molar basis). At this ratio, Altshuler
obtained about 0. 25 ppm "oxidant" (ozone) and 15% hydrocarbon consumption
after six hours irradiation.
Irradiation conditions for the butane runs were the same as for the propylene
runs, (95°F, 25% relative humidity, approximately 10% initial NO2 in NOx)
but the system was run for six hours, rather than five. Two materials,
Teflon and aluminum, were tested at two S/V values each. The single cleaning
technique of purge cleaning was used.
Results are summarized in Table 3-10. (Appendix A gives the run graphs and
the run parameter calculation listing. ) The full spectrum results were com-
parable to those of Altshuler. The average butane consumption was 16%,
considering the dilution correction, which was made individually for each run.
The average peak ozone concentration was 0. 23 ppm, and the ozone con-
centration appeared to be still slowly increasing at the six hour end-of-run
time. Time to NO- maximum averaged 253 minutes. S/V ratio had major
effects on NC>2 Max Time, % HC Disappearance and for Teflon on Ozone Max.
Cone.
3-22
LOCKHEED MISSILES & SPACE COMPANY. INC.
-------
LMSC-D406484
TABLE 3-10
Butane Effects by Materials*
Effect
1 N0_ Rate 2 NO. T
i i max
3 NO2 Dose 18 Dose Factor
Alum Teflon Alum Teflon Alum
m
A
B
AB
Effect
m
A
B
AB
Effect
m
A
B
AB
Effect
m
A
B
AB
2. 11 2
.43
1. 33
-.47
.00 125
.25 -1.5
.75 19.5
.01 -1.5
20 Crossover Time
Alum
132
-38
-54
3
4 Ozone
Alum
.72
-.05
1. 38
-.12
19 Ozone
Alum
5. 1
1.6
7.9
.2
Teflon
127
-22
-42
-6
Max Rate 5
Teflon
.84
.61
1.08
.01
Dose Factor
Teflon
5.5
1.8
5.0
4. 1
267
-95
-50
40
22 =
Alum
179
31
-43
-10
138
28
40
-3
2-21
Teflon
141
-73
-8
46
Ozone Avg Rate
Alum
. 19
-.03
.36
-.04
Teflon
.26
.24
.42
. 14
13 HC Max Rate
Alum
1.9
.05
.53
-. 60
Teflon
2.2
2.7
1.6
-1. 1
Teflon Alum
132 63
6 13
44 20
9 -1
23 =360-21
Teflon
61
5
18
5
17 NO Rate
Alum Teflon Alum
228 233 1
38 22
55 42
-36-
.50
. 33
.60
. 08
6 Max Ozone Cone.
Alum Teflon
, . 14 .13
.02 .07
.20 .18
-.02 .03
14 HC Avg Rate
Alum Teflon
1.0 .84
.99 -.11
.32 .69
.21 -1.1
Teflon
1.52
-. 11
.09
. 32
8 Ozone Dose
Alum
11
3
18
1
Teflon
12
5
12
9
% HC Disappearance*
Alum
14
10. 3
-1.4
-.3
Teflon
13
2
7
-11
*Those parameters that are inappropriate or poorly defined are omitted from
this table, i. e. N©2 FWHM, Max PAN concentration, Max. Aid. Cone. , HCT
HCT . 5, HCT . 25. Ozone Tmax was 360 minutes for all runs. Rather than
tabulate HC Final Cone. , the loss of butane - after dilution correction - is
tabulated as % HC Disappearance.
75,
3-23
LOCKHEED MISSILES & SPACE COMPANY. INC.
-------
LMSC-D406484
The cut spectrum drastically lowered ozone production (the average ozone maximum
concentration for the cut spectrum condition was 0.04 ppm) and increased the time to
NO2 maximum to an average of 328 minutes. Conclusions drawn from this small num-
ber of tests are somewhat tentative but are stated here: (1) the cut spectrum condition
has the same effect in butane/NO as in propylene/NO of slowing the photochemical
X X
reaction system (2) S/V remains as an important variable in affecting the photo-
chemistry; (3) measurable amounts of acetaldehyde (about 0.2 ppm) and PAN (about
3 ppb) were detected in two of the eight runs. These were both the high S/V aluminum
runs.
3.7 Background Reactivity Runs
For each material at high S/V, background reactivity tests were made, both at the
initial use of the material, and at the end of the material block investigation. Conditions
for these tests were standard except that the chamber charge was 0.1 ppm of NO and
X
no hydrocarbon was charged. The average maximum ozone concentration for those
tests was less than 0.02 ppm, and the NO0 formation rate was about 0.2 ppb/min.
£t
There was little observable difference between the results before or after the block of
photochemical runs.
3.8 "Virgin Surface" Effect
For the metallic materials, the initial photochemical run gave a different kinetic picture
than did the second or third replicate at the same constant condition. These initial runs
have not been used in the data analysis. For aluminum, the "fresh" surface gave a faster
reaction than did the sebsequent runs. For stainless steel, the initial run was slower
than the subsequent runs. These effects have not been analyzed in any detail, but are
present.
The "fresh" surfaces referred to had been through the standard cleaning technique of ace-
tone or trichloroethylene cleaning, followed by several distilled water rinses, followed by
overnight or longer hold at 2 microns or less pressure at about 100° F. Then the ozone
decay studies were performed. This involved exposure to 1 to 3 ppm of ozone for usually
3-24
LOCKHEED MISSILES & SPACE COMPANY. INC.
-------
LMSC-D406484
3 days continuously. Then the background reactivity run was performed, which involved
irradiation with 0.1 ppm NO in the system. Evidently, metallic smog chamber sur-
A.
faces do not stabilize until being exposed to at least one full scale photochemical run.
Cleaning as performed by this project did not degrade the stability of the Teflon or
aluminum surfaces. Cleaning appears to be affecting the stainless steel surfaces, and
may be affecting the Pyrex.
3.9 NITROGEN BALANCE
It is commonly believed that prior to NO0 peak, NO and NO0 are the only nitrogen com-
£i £
pounds present in the photochemical reaction system at measurable concentrations (say
greater than 0.05 ppm). In fact only recently has the chemiluminescent method made
direct measurement of NO possible. In this project, the Saltzman NO_ (after accounting
£
for instrument lag time) and the chemiluminescent NO are taken as correct measure-
ments. Using this ground rule, the NO, NO9, NO stoichiometry can be studied. The
^ X
ratio of NO0 at peak NO0 to initial NO is shown in Table 3-11. It is observed that the
£ £t X
ratio is significantly affected by S/V for Pyrex.
Such "excess NO2" behavior has been observed by other smog chamber operators but to
our knowledge has not been discussed in the literature. It seems to occur with "fast"
photochemical systems. The behavior may indicate that the Saltzman NO_ readings in-
^
eludes a response to a species other than NO0.
Z
Another disconcerting observation is the apparent non-stoichiometric behavior of NO0
£t
and NO during the first few minutes of irradiation. In this interval NO2 builds up while
NO decreases at a lesser rate. This induction period effect causes parameter 17 (NO
disappearance rate) to be smaller than parameter 1, the NO0 formation rate. Table 3-12
£t
gives the ratio of NO0 rate to NO rate. Material does significantly affect this ratio, in
£
the sense that the fastest photochemical material-stainless steel - gives the highest
ratio. The ratio is sensitive to cleaning technique for all the materials studied, and is
also sensitive to S/V for stainless steel.
3-25
LOCKHEED MISSILES & SPACE COMPANY. INC.
-------
LMSC-D406484
Table 3-11
NO at Maximum/NO Initially
£ X
Material
Run
Condition
+++
+ + -
-+ +
-+-
+ -+
+ --
--+
_ _ _
Stainless
Steel
.99
1.05
.99
1.00
1.05
.98
.93
.95
Alum.
1.08
1.05
1.08
1. 07
1. 08
.99
.98
.95
Pyrex
1.02
1. 08
.86
.92
1.03
1.19
.81
.83
Teflon
1.05
1.03
1. 12
1. 15
.93
1.00
1.05
1. 13
Averages
All Runs
S/V High
Low
Spectrum Full
Cut
Cleaning
Vacuum
Purge
1.00
1.02
.97
1.01
.99
.99
1. 00
1.04
1. 05
1.02
1.07
1. 00
1.06
1. 02
.97
1.08
.86
.97
.97
.93
1. 00
1.06
1. 00
1. 11
1. 09
1.03
1. 04
1. 08
3-26
LOCKHEED MISSILES & SPACE COMPANY. INC.
-------
LMSC-D406484
Table 3-12
NO2 Rate Divided by NO Rate
Run
Condition
+ ++
+ + -
-++
-+-
+ -+
+ --
--+
_ _ .
Stainless
Steel
1.74
2. 35
1.24
1. 61
1.71
2.03
1.20
1.77
Alum.
1. 39
1.64
1. 44
1.68
1.28
1.70
1. 31
1.39
Pyrex
1.24
1.39
1.06
1. 10
1.23
1.45
1.07
1. 04
Teflon
1.01
1.28
1. 19
1. 35
1.05
1.22
1. 16
1.40
Averages
All Runs
Full Spectrum
Cut Spectrum
High S/V
Low S/V
Cleaning
Vacuum
Purge
1.V1
1.74
1.68
1.96
1.46
1.47
1.94
1.48
1.54
1.42
1.50
1.46
1. 36
1. 60
1..20
1.20
1. 20
1.33
1.07
1. 15
1.25
1.22
1.25
1.20
1. 14
1.28
1. 10
1. 31
3-27
LOCKHEED MISSILES & SPACE COMPANY. INC.
-------
LMSC-D406484
It is generally acknowledged that nitrogen balance in smog chambers is somewhat
unsatisfactory. We observe that this lack of balance is not restricted to the post-
NO2 peak time, (at which time nitric acid formation is usually invoked), but includes
the early period. An additional complication in interpreting this data is the effect of
the delay time between sampling the reactants in the illuminated chamber and analysis.
The rapid reaction between 03 and NO in the dark to form NO2 reduces the apparent
concentrations of these two species. Instrument response times also affect the indi-
cated concentrations. The NO2 instrument has a rise time of about 10 minutes, while
for the chemi-luminescent NO and Og instruments, the rise times are about 3 seconds.
The NO2 readings were corrected by subtracting 10 minutes from the time of each
reading. This compensates for the time lag while ignoring the integration effect. The
dark time before sampling has been calculated at about 9.6 seconds. For an 03, NO
rate constant of 29.5 ppm"1 min'1, kt is equal to 4.46 ppm . At this value, 03 and
NO would react and not co-exist at concentrations above about 0.1 ppm. It thus
appears that the calculated dark time is too high or that the 03 and NO instruments
give high readings. The latter hypothesis is rejected on the basis of (1) pre- and post-
run zero checks frequently, showing zeros within 0. 02 ppm of the anticipated 0.00
value; (2) experiments in which excess NO was added to the chamber at the end of
the photochemical run (lights off) and 03 was recorded at about 0. 02 ppm (accuracy of
Og meter zero confirmed); and (3) the normal check reading at the end of each photo-
chemical run (lights off) which gives a condition of excess Og in the chamber and which
did give an NO reading of about 0.02 ppm (accuracy of NO meter zero confirmed).
Additional effort in NO and NO2 monitoring techniques, such as gas-phase titration of
NO with Og (thus establishing NO by the decrease in 03), and in linking NO2 and NO
monitoring by using the NOX mode of a chemiluminescent detector would be useful in
resolving these problems. It is believed that to start resolving these questions,
monitoring of the nitrogen species also must be accomplished by a technique sensitive
to more than NO and NO2. Such monitoring may be in-chamber or with a few seconds
sampling time delay. Two promising methods are long path length IR or Fourier
Transform Spectrometry, or by time derivative spectroscopy. Either seem well
suited.
3-28
LOCKHEED MISSILES & SPACE COMPANY. INC.
-------
LMSC-D406484
Section 4
DISCUSSION
4.1 General Observations
The large effect of spectral distribution on photochemical reaction systems - at the
same intensity as measured by k, - has not previously been reported. However, some
literature data are available that suggest that such an effect is not only present but
may be general. Altshuler and Cohen (Ref. 5) reported a factor of 2 to 3 times higher
NO formation rates for tests in Teflon vs. Mylar containers. This difference was
-1 -1
attributed to the difference in k, of 0.35 to 0.4 min for Teflon vs. 0.25 to 0.3 min
d
for the Mylar. The substantial difference in light below 330 nm for the two materials
was noted, but not further discussed. This differential rate was observed for some 16
hydrocarbons, ranging in reactivity from 1,3,5 trimethylbenzene and 1, 2, 3, 5
tetramethylbenzene at the high reactivity end, to ethylbenzene and toluene at the low
reactivity end.
Table 4-1 gives a "Mylar/Teflon" spectral effect for time to NO2 maximum calculated
from Altshuler's data, after normalizing by the ratio of 0.275/0.325 to account for
4-1
LOCKHEED MISSILES & SPACE COMPANY, INC.
-------
LMSC-D406484
the difference in k . Glasson and Tuesday (Ref. 6) give experimental data that show
that NO- formation rate is linear with k, for a variety of hydrocarbons. Such a linear
factor is also suggested by Niki et al. (Ref. 7) as applicable to the early stages of the
photochemical reaction. The difference in transmission characteristics for the Teflon
vs. the Mylar containers results in a cut/full spectral distribution somewhat similar to
that used in the present study. The spectral effect factor for propylene was 1.65.
A similar treatment of that data for oxygenates yields the following "Mylar/Teflon"
spectral effect: formaldehyde 5.9, acetaldehyde 3.5, proprionaldehyde 3.9, acrolein
2.9, ethanol 1.6.
Table 4-1
"MYLAR/TEFLON" SPECTRAL EFFECT*
Hydrocarbon Ratio of NO_ T
—* 2—max
Ethylene 2.93
Propylene 1.65
Isobutene 1.81
Toluene 1.76
Ethylbenzene >2.7
1,2-dimethylbenzene 1.46
1,3-dimethylbenzene 1.32
1,4-dimethylbenzene 2.20
1,2 -methylethylbenzene 1.74
1,3-methylethylbenzene 1.96
1,4-methylethylbenzene 2.29
1,3,5-trimethylbenzene 2.00
1,2-diethylbenzene 2.05
1,3-diethylbenzene 1.67
1,4-diethylbenzene 2.06
1,2,3,5-tetramethylbenzene 1.92
* Calculated from Ref. 5,
4-2
LOCKHEED MISSILES & SPACE COMPANY. INC.
-------
LMSC-D406484
Bufalini et al. (Ref. 8) have reported that photooxidation of formaldehyde in the pres-
ence of NCL proceeded more rapidly at k, of 0.14 min~ with sunlamps than at 0.32
-1
min with blacklamps. This is an obvious result of the photodisintegration by the
shorter wavelength light from the sunlamps. The time ratio for 50 percent consump-
tion of formaldehyde, corrected for the k, ratio, is 1.6 for the blacklamp/sunlamp
distribution. For an irradiation without NO2, the corrected time ratio for 37 percent
consumption is 2.7.
4.2 NO2 Photolysis Distribution
The distribution of NO2 photolysis events versus wavelength has been calculated for
several spectral distributions. For this calculation, the wavelength interval between
290 and 410 nm is considered. Absorption coefficient (Ref. 9) and quantum yield (Ref. 10)
for NO2 are tabulated at 10-nm intervals in this interval. These multiplied by each
other and by the number of photons in the 10-nm wavelength interval gives the total NO2
photolysis rate. This total rate divided into the events in each 10-nm band gives the
fractional distribution of NCL photolysis events, or shows how the same k, occurs for
different spectra. Figure 4-1 gives the results.
!/>
g
5
o
o
O
5
CM
O
O
0.2fi
p
p
o>
p
S
o
8
SUNLIGHT
XENON-FULL
XENON-CUT
NAPCA CHAMBER
1
/
»
280 300 320
340 360 360 400
WAVELENGTH (nm)
420
Fig. 4-1 Distribution of NCL Photodisintegrations for Various Spectra
4-3
LOCKHEED MISSILES & SPACE COMPANY. INC.
-------
LMSC-D406484
The sunlight curve is for high noon in Los Angeles or Stanford, Conn, (two distribu-
tions averaged) (Refs. 11, 12); the xenon-full and xenon-cut are for the spectra used in
this study, as previously shown in Fig. 2-6. The NAPCA chamber curve is for the
spectral distribution given by Korth et al. (Ref. 13), which is one of the few reported
chamber spectra in the literature.
The xenon-full NO2 disintegration spectral distribution does indeed closely match the
sunlight curve, over the entire wavelength interval. The NAPCA chamber distribution
is probably typical of chambers illuminated by fluorescent tube combinations, and shows
that the wavelength band from 335 to 365 nm is overemphasized (relative to the sunlight
distribution), while the 385-410 nm band is underrepresented. This disparity is fre-
quently unrecognized in discussing the application of smog chamber data to atmospheric
simulation.
It is clear that the spectral distribution of light provided a smog chamber influences
the photochemical reaction observed. In particular, the NO2 photodisintegration rate
(k, or its equivalent k-) does not sufficiently characterize the light conditions.
4.3 Possible Mechanisms
There are two major effects demonstrated by the experimental work that are not well
accounted for in present-day photochemical reaction system kinetic modeling studies.
These are: (1) the spectral effect and (2) the material and S/V effects (in fact, cham-
ber wall effects usually are not included in such models.) An effort is presently
underway to examine the propylene/NO system reactions by an updated kinetic model
A
and to determine those mechanisms that best suit the experimental findings. This
analysis is being submitted for publication to Environmental Science and Technology
as "Interpretations of Smog Chamber Design Effects," P. S. Connell, R. J. Jaffe,
and H. S. Johnston. The model accommodates up to 200 elementary reactions and 50
species, and is intended to include all important elementary reactions. It is well known
that there are a number of reactions in the system that are spectrally sensitive.
These are listed as follows.
4-4
LOCKHEED MISSILES & SPACE COMPANY. INC.
-------
LMSC-D406484
Effective
Wavelength (nm)
(1) HONO + hv *-HO + NO 300-390
(2) HCHO + hv ^H + HCO 290-360
(2a) HCHO+hv *> H2 + CO 300-370
(3) CH-CHO+hi; *>CH0+HCO 300-350
o o
(3a) CHg CHO + hv » CH4 + CO 300-350
(4) H2O2+hv ^2 HO 300-370
(5) 03+hv *>02(lAg) +0(1D) 290-310
(5a) O (1D) + H0O ^ 2 HO
Lt
The mechanism and rate constants for which no experimental evidence exists are drawn
largely from the modelling work of Demerjian, Kerr, and Calvert; and Niki, Daby and
Weinstock. When available, CIAP rate constants were used as well as more recent
experimental work. The photolytic constants for nitrous acid and for the production of
O( D) were developed from recent studies at the University of California of cross sec-
tion and quantum yield by R. Graham, R. A. Cox (nitrous acid) and J. Girman (un-
published study of O( D) production).
For the full spectrum data, the sunlight cross sections and yields have been used, after
normalizing for intensity by the k, ratios. The cut spectrum distribution and photolytic
data have been used to assign new rate constants for the above reactions. Nothing
else in the model was changed. The resulting comparison of cut/full spectrum con-
ditions gives a ratio for time to NO2 maximum of 1.37 to 1.81, depending upon the
initial nitrous acid concentration. The midpoint of about 1.6 agrees well with the
experimental data from this project, and from Altshuler's 1963 work.
The responsible reactions were identified by replacing individual full-spectrum 'j1
values with the corresponding cut-spectrum values, one by one. The largest effect
was observed for nitrous acid, followed by, in order, formaldehyde, acetaldehyde,
and hydrogen peroxide. The production of O( D) in the photolysis of ozone appears to
be unimportant in the smog chamber. Table 4-2 gives the ratio of the cut spectrum to
full spectrum time to NO2 maximum considering the change for each species, individually.
4-5
LOCKHEED MISSILES & SPACE COMPANY. INC.
-------
LMSC-D406484
Table 4-2
SPECTRAL EFFECT ON TIME TO NO£ MAXIMUM
CAUSED BY EACH SPECIES
Species Ratio of Times for Cut/Full Spectrum
Nitrous Acid 1.50
Formaldehyde 1.26
Acetaldehyde 1.17
Hydrogen Peroxide 1.12
Og — O^D) 1.00
The observed surface effect is that 'active1 surfaces, such as stainless steel, speed
the smog reaction in comparison to inert surfaces, such as Teflon. The net contribu-
tion of heterogeneous reactions must not be in the direction of radical quenching but
catalysis of initiating or chain propagating processes. By examination of propene
destruction data as a function of photons absorbed by NO2, it was found that the stain-
less steel effect is one of increased initial effective quantum yield. The important
identified heterogeneous reactions are (1) the formation of nitric acid by the reaction
of N0O_ with absorbed water and (2) the wall dependent formation of nitrous acid.
£t O
Unpublished work in two cells, one of quartz and the other of quartz and stainless
steel, shows that the stainless steel surface promotes the rate of formation of nitrous
acid by a factor of about a thousand (as observed in the gas phase). Since the photolysis
of nitrous acid has been seen to be an important initiator, its rate of formation and
concentration at the beginning of the irradiation time are significant parameters in
determining the overall rate of the smog reaction.
4.4 Background Reactivity Runs
The background reactivity runs were performed to determine whether the cleaning
techniques had removed trace amounts of reactive species. Tests were conducted, in
which about 0.1 ppm of NO was charged to the chamber and NO, NO0 and OQ monitored.
X £ o
The results are summarized in Table 4-3.
4-6
LOCKHEED MISSILES & SPACE COMPANY. INC.
-------
LMSC-D406484
Table 4-3
BACKGROUND REACTIVITY RESULTS
03 Maximum NO2 Formation Rate
Material and Conditions (ppm) (ppb/min)
Aluminum - Post Runs - High S/V 0.012 0.2
Aluminum - Post Runs - Low S/V 0.014 0.2
Pyrex - Pre Runs - High S/V 0. 024 0.3
Pyrex - Post Runs - High S/V 0.012 0.2
Teflon Pre Runs - High S/V 0.014 0.3
Teflon Post Runs - High S/V 0.020 0.2
The NO2 formation rate of 0.2 to 0.3 ppb/min agrees well with the reported rate for
similar background runs conducted by the University of North Carolina using a
3
6000 ft Teflon outdoor smog chamber (Ref. 17). For those tests the NO charge was
0.2 ppm, contrasted to the 0.1 ppm charge used with this chamber. Thus this 100-fold
smaller chamber has a background NO2 formation rate about 4 times higher. There
is no indication of residual reactive species upon comparing the tests conducted at the
start and at the end of a test series.
4.5 Ozone Decay Results
The ozone decay results indicate that the stability of ozone in the presence of the vari-
ous materials is greatest for pyrex, followed in order by teflon, aluminum, and stain-
less steel. Ozone decay time in the light is 0.4 to 0.8 of the value measured in the
dark. Ozone decay rate is usually considered to be a measure of the cleanliness of the
chamber. Recently Dodge and Hecht (Ref. 18) analyzed ozone decay data from the Uni-
versity of California, Riverside evacuable smog chamber. Ozone half life in the dark
was 501 min, and in the light 260 min. The difference in half lives could be accounted
for by ozone photolysis, so the conclusion was that stability of ozone in the light is not
a good measure of chamber contamination. Conditioning effects do occur in ozone half
life determination. Such effects were observed in this study and have been reported by
Sabersky (Ref. 19). Such conditioning effects probably reflect removel of surface con-
taminants by the ozone.
4-7
LOCKHEED MISSILES & SPACE COMPANY. INC.
-------
LMSC-D406484
Section 5
RECONCILIATION OF CAPI-6 DATA
The Coordinating Research Council Project CAPI-6 (Techniques for Irradiation Cham-
ber Studies) generated data from some ten smog chambers. A group of round-robin
tests were conducted using these ten chambers as follows: (1) irradiation of seven
different hydrocarbons with nitrogen oxide; (2) replicate runs to establish reproducibil-
ity using the propylene-nitrogen oxide system, and (3) a reactant concentration study
in which 3 ppm propylene was reacted with 3, 1.5, and 0.5 ppm nitrogen oxides (Ref.
2). Table 5-1 gives the chamber characteristics. The wide variation in chamber
size, materials, lighting conditions and type of lights is notable.
It is not surprising that these 10 chambers should give different values for the photo-
chemical run parameters. The run parameters reported for the 3 ppm propylene/
1.5 ppm NO system are given by Table 5-2. The variation between chambers was
about a factor of two or three, as shown below:
Parameter Minimum (Laboratory) Maximum (Laboratory)
NO0 Formation Rate (ppb/min) 18 (A,F, J) 42 (C,D)
£i
Time to NO9 Max. (min) 30 (I) 73 (A)
u
NO2 Dose (ppm-min.) 76 (H, J) 296 (E)
Oxidant Max. Cone, (ppm) .46 (D) 1.4 (C)
Time to Oxidant Max. (min) 90 (G) >360 (E)
Oxidant Dosage (ppm-min) 57 (D) 240 (H)
Propylene Final Cone, (ppm) . 01 (B, F, G, H, I) 0.66 (E)
Propylene Half Time (min) 54 (B) 104 (A, E)
Propylene Max. Rate (ppb/min) 12 (E) 41 (B)
The analysis described below has shown that the observed differences in chamber behavior
can largely be accounted for by variations in chamber lighting, chamber materials S/V,
chamber volume, and initial conditions. To obtain this normalization of chamber
behavior, a combined photochemical and statistical approach was used, in which the
experimental data from the factorial experiment was used along with the CAPI data.
This provided a data base with sufficient variation in materials and k, effects to obtain
5-1
LOCKHEED MISSILES & SPACE COMPANY. INC.
-------
298
450
1.5
335
330
1.0
6.7
22.4
3-35
14
65.8
4.6
2.9
14.2
4.9
116
14?
l
.25
ll4o
1300
1.
15
64
92
1.44
1.8
T.I
3-9
6lO
479
0.78
Table 5-1 Chamber Characteristics
A B C D E F G H
Volume, ft3 2
Surface Area, ft
S/V, ft.'1
0 Surface Material, y>:
n Stainless Steel 67 53 56
I Aluminum 45 43 64 79
m Nickel 90
g Glass &4 29 57
_ Pyrex 33 10 32 99
± Teflon 16 18 1 21
$ Tedlar 55 1^
Uj Light Intensity-kd .37 -4 .4 .42 .16 .36 .6* .49 .2 .3**
W Lighting System:
g> Internal X XX X
y, " External XX X XXX
TJ Black Lights X X XX X X X X X X
£ Blue Lights XX X XX
m Sunlamps XX XX
O Ozone Decomposition:
O T 1/2, dark, hr. 10 10*** 18.5 3.5 NA 6 7 12 56 8
| T 1/2, light, hr. 2.2 7*** 5-5 0.85 NA 1.5 4 39 1.5
Z
NA = Not Available
O * Private communication, J. Shikiya to R. J. Jaffe, 4-25-75.
** Calculated from data supplied in private communication, D. Miller to R. J. Jaffe, 4/11/75*
*** Calculated from reported loss .in 3 1/2 hours.
-------
Table 5-2 CRC-APRAC Irradiation Chamber Comparison - Study III
Propylene Reactivity Comparison: 3 PPm Propylene - 1.5 ppm NO
Laboratory . • A B C D E F G H_
Initial Concentrations
0 Propylene, ppm
O NOX, ppm
j N02 , ppm
m
g NOp Formation
2 NOp Formation Rate, ppb/min
55 ' Time to NOp Maximum, min
52 N00 Dosage, ppm-min
r *=
m
U) Oxidant Formation
* Maximum Rate, ppb/min.
U) Average Rate, ppb/min. '
J! Maximum Concentration, ppm
O Time to Maximum, min.
m Oxidant Dosage, ppm-min
O
2 Propylene Disappearance
> Final Concentration, ppm
T0 75'
T min
3.0
1.53
0.12
18
73
118
22
6
1.1
135
195
0.11
79
10U
i4o
25
l^ '
3.06
1.U9
0.09
• 33
35
1^9*
37-
12.
1-3
93
173*
0.01
39
54 '
78
Ui
28
2.0
0.28
3.10
1.86
0.38
U2
34
210
20
13
1.4
175
317
0.02
29
55
76
33.
29
3.15
1.4-9
0.13
42
32
110
13
5-
0.46
82
57
o.4o
30 •
71
118
17
20
1.5
3.03
1.50
0.18
20
85
296*
2.5
2.0
0.7
>36o*
103
0.66
56
104
159
12
11
2.94
1.55
o.i4
18
71
159
16
6
1.0
>300*
203
0.01*
61
90
136
21
18
1.5
3.0
1-5
0.16
28
50
109*
27
7
1.0
90
185*
0.01
48
69
95
31
21
3.0
1.5
0.06
23
56
76
42
9
1.2
106
240
0.01
44
68
89
35
22
2.4
0.47
3.0
1.5
0.06
38
30
87
25*
11
0.8
124
65
0.01*
31
60
88
26
25
2.8
3.
l.
0.
18
58
77
27
5-
0.
93
114
0.
74
92
113
38
16
0,
05
60
03
8
8
10
• 9
.,
Maximum Rate, ppb/min
Average Rate, ppb/min
Maximum CH^O, ppm <=«u J-O 1O 2 .4 2.O 0.9 tf
Maximum PM,ppm 0.28 . O.k o
values vere determined by personal communication with the experimenter or calculated from other
tests conducted for CAPI-6.
-------
LMSC-D406484
2
model equations for the various run parameters that are high enough in R that they
are believed to be realistic.
5.1 CAPI-6 DATA HANDLING
The seven materials shown in Table 5-1 were reduced to four by grouping similar
materials. Stainless steel was used for both stainless steel and nickel; Pyrex was
used for both Pyrex and glass; and Teflon was used for both Teflon and Tedlar. These
groupings were made in consideration of the known similarities in surface properties
of the materials, which leads to the expectation that effects of the paired materials
are close enough to justify the pairings. Cleaning technique used for the CAPI-6
chambers was not systematically varied. Some chambers used overnight purges at
elevated temperature, and some used vacuum offgassing techniques. It was thus not
possible to treat cleaning as an independent variable.
Light intensity for the chambers was reported as k,. The types of lamps used were
also reported. Basically two systems were used. Chambers A, B, F, H, and J used
a combination of sunlamps, black lamps and blue lamps. The mix used for the lamps
varied somewhat from chamber to chamber. The other chambers used black lights
only. The factorial experiment results showed that light at wavelengths below 340 nm
was proportionally more important in speeding the smog rea ction than was light at
higher wavelength. To account for this effect, the regressions were carried out to
group chambers A, B, F, H, and J and chambers C D, E, G, and I as having two
separate k, effects. That is, it was assumed that the combination-lamp chambers
had the same spectral effectivity and that the black lamp chambers could have a dif-
ferent spectral effectivity.
5.2 FACTORIAL EXPERIMENT DATA HANDLING
The factorial experiment, as described in Sections 2 and 3, resulted in data for the
photochemical run parameters for a total of 32 chambers, as obtained from four
materials times 2 S/V values times 2 spectral conditions times 2 cleaning conditions.
5-4
LOCKHEED MISSILES & SPACE COMPANY. INC.
-------
LMSC-D406484
The full spectrum condition is the one that is closest to the CAPI spectral distribution
and was the one used. As the CAPI data was gathered from chambers that used a
variety of cleaning conditions, the average of the two cleaning techniques was selected
from the factorial experiment results. This gave run parameters for eight chambers.
Data for a ninth chamber was also available. The ninth chamber was the baseline,
empty Pyrex chamber with no added surface materials. The photochemical runs for
all these chambers were conducted at k, of 0.3 min .
d
5.3 MULTIPLE REGRESSION RESULTS
The nine factorial experiment runs and the ten CAPI runs formed the 19 point array
analyzed in the multiple regression. Twelve independent variables have been included
in the regression model:
Aluminum S/V, ft"1
Stainless Steel S/V, ft"1
Pyrex S/V, ft"1
Teflon S/V, ft"1
Q
Chamber Volume, ft
Initial HC, ppm
NO ppm
X
NO2, ppm
% N02 in N0x
HC/NO Ratio
-1
k, ACT, min
d -1
k, ADJ, min
These were selected based on the results obtained in Appendix B, for correlating the
factorial experiment; and on the previously discussed basis that k, and spectrum com-
bined would give a better fit than a single k,. Thus k, is entered in two ways: k, ACT
is the actual k, for the factorial experiment runs and the CAP!chambers A, B, F,
H, and J but zero for the remaining CAPI chambers; and k, ADJ is the actual k, for
CAPI chambers C, D, E, G, and I, and,zero for the remaining chambers.
5-5
LOCKHEED MISSILES & SPACE COMPANY. INC.
-------
LMSC-D406484
Running multiple regressions (using the BMD 03R multiple regression program devel-
oped at the University of California at Los Angeles) with the twelve independent varia-
2
bles, the prediction equations and R values for the nine run parameters considered
2
are as found in Table 5-3. It should be noted that the R value is the percentage of
the observed variability which is accounted for by the independent variables, and
2
ranges from 75.5% to 89.7%. While the R values are not particularly high when con-
sidering the number of data points and the number of independent variables, it is
encouraging that they are as high as they are. Tables 5-4, 5-5, and 5-6 give the re-
sults of the multiple regression prediction equation; the residuals, and the normalized
residuals. The normalized residual (Table 5-6) is the residual divided by the respective
parameter standard deviation. For 16 of the 19 chambers, the normalized residual
is less than 3, indicating satisfactory fit of the data. It should also be noted that the
normalized residuals are well distributed among the 19 chambers, with the possible
exceptions of oxidant maximum concentration and oxidant dose, for which the
factorial experiment results form extremes.
To find whether the double entry of k, (to allow for spectral differences) was advanta-
geous, the full model was changed so as to include only a single k, term, that term
being the actual k, for each chamber. The 10 other variables were unchanged.
2
Table 5-7 gives the R comparisons of the dual and single k, models.
2
The R values show a decrease from the dual k, model to the single k, model for all
parameters except those dealing with NO0, and the HCT 50, indicating that the dual k,
^ d
allowing for a spectral effect is beneficial in the parameters. Through the decrease
is small for some of the remaining parameters (notably O0TM), it is expected that the
2
decrease in R is caused by the removal of a needed term, rather than merely because
2
fewer independent variables were used (the drop in R , with the possible exception of
(XTM, is large when considering that the number of independent variables is being
o
decreased by only one variable, or by 8.3%). The insensitivity of the NO2 parameters
to use of a second k, is not surprising, as k, is a measure of NCL photolysis rate, and
the regression equation fit can be satisfied by a single value of k,.
5-6
LOCKHEED MISSILES & SPACE COMPANY. INC.
-------
Table 5-3 Results of Multiple Regression with Full Model
1
1/N02R
r
0
O
*
I .
m
m
0
2
0)
en
r
m
(A
8s
U)
T)
o
m
O
0
TJ
2
Z
o
f
VOL
ALUM
SS
FYR
TEF
kdACT
HC INZT
NO
X
N02
$NO
HC/NOX
kdADJ
CONST.
77.1
-0.00001
-0.00847
-0.01019
-0.00488
0.01233
-0.07332
0.45600
-0.61969
-1.65762
0.02811
-0.72429
-0.07619
1.07049
2
81.7
-0.01601
-7.27405
-5.25568
0.74252
9.04682
-32.96704
306.51807
-321.92065
-2985.35156
49.98193
-642.34033
-48.49976
887.02930
3
81.9
-0.01544
-20.19029
-10.46144
-29.32697
0.80514
-454.05908
541.18042
-1205.9399^
1&57.54834
-19.47304
-746.3879^
-354.90674
1945.75562
Run Parameters
6 7
0-MAX C 0-TM
79-5
0.00026
0.01021
-0.06142
-0.0:558
-0.01325
0.40104
-4.11157
3.86622
20.30347
-0.29726
5.79683
-0.50213
-4.14520
75.5
-0.13341
-42.o84l4
-0.47079
-36.58080
-16.99710
-367.24854
1887.50269
-3732.97095
-1741.55469
33.22971
-3714.80566
-167.00391
7652.95703
8
89.7
0.01296
11.49137
4.317^5
2.44492
-8.28517
383.83301
-1035.93994
1284.81372
2178.25098
-28.72931
1124.15088
169.23122
-1055.64429
9
HCFC
80.8.
-0.00022
-0.03204
0.03437
-0.04372
0.00475
-0.70153
2.20197
-1.87708
-10.29309
0.16763
-2.73319
-0.17421
1.94828
11 13
HCT50 1/HC MAX RATE
79
-0
-10
.1
.01981
.85161
-8.74458
-1
10
-82
450
-474
-3070
51
-764
-76
976
.72086
.06901
.51172
.18188
.81421
.67969
.20485
.72607
.15112
.24683
79 -^
-0.00004
-0.00890
-0.00152
-0.00808
-0.00032
-0.07654
0.37821
-0.59283
-0.85324
0.01431
-0.59400
0.00219
1.01017
Ul
-q
o
-------
Table 5-4 Run Parameters Predicted From Multiple Regression
Run Parameters
CHAMBER
A
5
O
I
m
m
o
z
ui
r
m
en
m
It SPACE
COMP,
f
Z
^
Z
p
B
C
D
E
F
G
H
I
J
ALUM HIGH
ALUM LOW
SS HIGH
SS LOW
TEF HIGH
TEG LOW
PYR LOW
PYR HIGH
PYR BASE
1/N02R
0.04834
0.02762
0.03124
0.03897
0.06735
0.03072
0.02641
0.02676
0.05841
0.06896
0.06726
0.03252
0.05560
0.09679
0.08533
0.06025
0.06835
0.05655
2
NOpTM
64.82300
44.87793
38.33400
39.53026
67.17046
88.22900
45.75050
36.34375
32.52100
62.74437
91.20044
90.33032
58.73291
82.96143
111.74316
109.72583
100.73438
87.68359
77.47168
NOgDOSE
134.57764
128.94653
218.17624
137.78392
252.26547
162 . 14966
96.66560
43.86914
83.43175
101.96338
171.86060
159.92212
154.96045
159.56519
176.66138
164.71191
124.99365
139.77075
171.57568
6
0-MAX C
1.07269
1.18542
1.38206
0.56036
0.63516
1.06523
0.97474
1.15695
0.75248
0.90182
1.01108
1.00162
0.97268
0.96508
1.11413
1.03559
1.10186
0.92530
1.19648
0 TM
178.80469
75.31641
193.38986
124.83102
306.89258
255.85547
66.25079
68.58984
110.87657
132.15234
159.29297
14 1.39063
126.53906
167.26953
179.79688
173.36719
157.36719
105.60156
159.61328
8
0 DOSE
194.35645
174
311
78
98
210
176
228
55
127
177
155
167
159
169
171
197
143
186
.26685
.96899
.74800
.14925
•57935
.73207
. 58740
.03642
.10132
.17993
.28076
.74487
.12183
.14233
.64917
.83032
.60522
.08130
HCFC
0.08407
0.00542
0.03499
0.39399
0.544o6
0.13546
0.01566
-0.08510
0.06837
0.08473
0.16738
0.16867
0.22869
0.21037
0.09115
0.15595
O.o44i4
0.15816
0.07348
11
HCT 50
86.34229
72 . 57080
59-17482
72.22748
93.41737
109,20288
66.97267
53.11987
63.61304
92.28516
114.23608
116. 46240
79.93945
107.34180
136.82568
135-88403
118.20264
120.57324
101.50000
1/HC MAX R
0.03662
0.02752
0.03338
0.064 11
0.07476
0.05041
0.02878
0.02109
0.03675
0.02855
0.04070
0.04363
0.03891
0.04654
0.04569
0.04661
0.03563
0.03444
0.04077
en
I
oo
-------
Table 5-5 Run Parameters Residuals from Multiple Regression
Run Parameters
r
0
O
I
m
m
o
2
55
en
r
m
en
fi»
en
TJ
O
m
O
o
2
TJ
Z
^
Z
o
CHAMBER
A
B
C
D
E
F
G
H
I
J
ALUM HIGH
ALUM LOW
SS HIGH
SS LOW
TEF HIGH
TEF LOW
PYR LOW
PYR HIGH
PYR BASE
1
1/N02R
-0.00743
+0.01&04
+0.00382
+0.00744
-0.01103
+0.01175
-0.00498
-0.01709
+0.00046
+0.00281
+0.00086
+0.00456
+0.00642
+0.00950
-0.00361
-0.01187
+0.01915
-0.00765
-0.02135
2
WOgTM
-8.17700
+9.87793
+4.33400
+7.53026
-17.82954
+17.22900
-4.24950
-19.65625
+2.52100
+4.74487
+1.20044
+7.33032
+12.73291
+16.96143
-0.25684
-18.27417
+10.73438
-9-31641
-17.52832
3
W02DOS
+16.57764
-20.05347
+8.17624
+27.78392
-42.73453
+3.14966
-12.33440
-32.13086
-3.56825
+24.96338
+9.86060
+1.92212
+26.96045
+22.56519
+8.66138
-35 .28809
+6.99365
-22.22925
+10.57568
6
0-MAX C
-0.02731
-0.10458
-0.01794
+0.10036
-0.06484
+0.06523
-0.02526
-0.04305
-0.04752
+0.10182
+0.05108
-0.07838
+0.10268
-0.19492
+0.04413
_0.i444l
+0.04 186
+0.01530
+0.22648
7
OJTM
•*3
-17
+18
+42
-53
-44
-23
-37
-13
+39
+10
+1
+21
+45
+2
-19
+25
-42
-0
.80469
.68359
.38986
.83102
.10742
.14453
.74921
.4ioi6
.12343
.15234
.29297
.39063
.53906
.26953
.79688
.63281
.36719
.39844
.36872
8
0 DOS
-0.64355
+1.26685
-5.03101
+21.74800
-4.85075
+7-57935
-7.26793
-ll.4i26o
-9.96358
+13.10132
+19.17993
-35.71924
+6.74487
-41.87817
-0.85767
-15.35083
+25.83032
+3.60522
+34.08130
9
HCFC
-0.02593
-0.00458
+0.01499
-0.00601
-0.1159^
+0.12546
+0.00566
-0.09510
+0.05837
-0.01527
+0.00738
+0.05867
+0.11869
+0.11037
-0.01885
-O.oi4o5
-0.06586
-0.05184
-0.08652
11
HCT 50
-17.65771
+18.57080
+4.17482
+1.22748
-10.58263
+19.20288
-2.02733
-14.88013
+3.61304
+0.28516
-2.76392
+10.46240
+7.939^5
+20.341&0
-2.17432
-14.11597
+12.20264
-6.42676
-27.50000
13
1/HC MAX R
-0.00338
+0.00352
+0.00338
+0.00511
-0.00824
+0.00241
-0.00322
-0.00791
-0.00125
+0.00555
-0.00130
+0.00563
+0.00391
+0.01154
+0.00169
-0.01139
+0.00763
-0.00556 K
-0.00823 "J
o
01
I
co
-------
Table 5-6 Normalized Residuals from Multiple Regression
Run Parameters
[—
o
o
I
D
2
u5
^
m
C/)
fi»
>
o
m
o
O
Z
5
^
2
n
CHAMBER
A
B
C
D
E
F
G
H
I
J
ALUM HIGH
ALUM LOW
SS HIGH
SS LOW
TEF HIGH
TEF LOW
PYR LOW
PYR HIGH
PYR BASE
AVERAGE OF
1
1/W02R
-1.11
+2.69
+ •57
+ 1.11
-1.65
+1.75
-.74
-2.55
+ .07
+ .42
+ .13
+ .68
+ .96
+1.42
-.54
-1-77
+2.86
-1.14
-3-19
1.33
2
N02TM
-1.14
+1.37
+ .60
+1.05
-2.48
+2.39
-.59
-2.73
+ .35
+ .66
+ .17
+1.02
+1.77
. +2.36
-.04
-2.54
+1.49
-1.29
-2.43
1.39
3
N02DOS
+2.72
-3.29
+1.34
+4.55
-7-01
+ .52
-2.02
-5-27
-.58
+4.09
+1.62
+ .32
+4.42
+3.70
+1.42
-5.78
+1.15
-3.64
+1.73
2.90
6
O..MAX C
-1.24
-4.75
-.82
+4.56
-2.95
+2.97
-1.15
-1.96
-2.l6
+4.63
+2.32
-3.56
+4.67
-8.86
+2.01
-6.56
+1.90
+ .70
+10.29
3.58
7
0-.TM
+5.62
-2.27
+2.36
+5-50
-6.82
-5.67
-3.05
-4.80
-1.68
+5.03
+1.32
+ .18
+2.76
+5.81
+ .36
-2.52
+3.26
-5.44
-.05
3.39
8
03DOS
-.10
+ .20
-.80
+3.45
-.77
+.1.20
-1.15
-1.81
-1.58
+2.08
+3.04
-5.67
+1.07
-6.65
-.14
-2.44
+4.10
+ .57
+5 .41
2.22
9
HCFC
-.74
-.13
+ .43
-.17
-3-31
+3.58
+ .16
-2.72
+1.67
-.44
+ .21
+1.68
+3-39
+3.15
-.54
-.40
-1.88
-1.48
-2.47
1.50
11
HCT 50
-2.43
+2.56
+ .58
+ .17
-1.46
+2.65
-.28
-2.05
+ .50
+ .04
+ .38
+1.44
+1.09
+2.80
-.30
-1.94
+1.68
-.89
-3.79
1.42
13
l/HC
MAX R
-.70
+ .73
+ .70
+1.06
-1.72
+ .50
-.67
-1.65
-.26
+1.16
-.27
+1.17
+ .81
+2.40
+ .35
-2.37
+1.59
-1.16
-1.71
1.11
Average
of
Absol. Value
1.76
2.00
•91
2.40
3.13
2.36
1.09
2.84
.98
2.06
1.05.
1.75
2.33
4.13
.63
2.93
2.21
1.81
3-45
2.09
£
§.
^
o
^
?
ABSOLUTE VALUE
en
i
-------
LMSC-D406484
Table 5-7
COMPARISON OF R2 FROM DUAL AND SINGLE kd MODELS
Dual k(j Model Single ty Model
Run Parameters (%)
1/NO2R 77.1 77.1
NO0 TM 81.7 81.6
£
NO2 DOS 81.9 80.2
OgMAXD 79.5 69.1
OgTM 75.5 70.7
O. DOSE 89.7 81.6
o
HC FC 80.8 74.0
HCT50 79.1 79.1
1/HC MAXR 79.4 61.8
To determine how good a fit could be obtained with fewer independent variables, the
Hocking-LaMotte regression algorithm was used. In employing this algorithm, k,
ACT and k, ADJ were forced into the model, and the remaining terms allowed to enter
2
as needed. The resulting range in R was from 60.1 to 86.3%. Table 5-8 gives the
results. Interesting items to note are that of nine possible times, % NO- enters eight
LI
times (accompanied by NO» seven of these times), that stainless steel S/V enters
five times, and that chamber volume enters only once. Another approach to a limited
number of variables data fit was to force into the model the two k, values, the four
2
material S/V values and the % NO0. This gave R of 51 to 79.5%.
64
From these multiple regressions, it is seen that most of the variability between the
chambers is due to some or all of the measured variables. Improvements might be
made by using cross products of various variables, or by using a more refined measure-
ment of the spectral effectiveness. These procedures were not attempted at this time
due to the relatively small number of data points in comparison with the cross products.
5-11
LOCKHEED MISSILES & SPACE COMPANY. INC.
-------
LOCKHEE
D
2
cn
F
m
»
TJ
n
m
n
O
2
TJ
Z
Z
n
i
1/NOPR
P
R 66.1
VOL
ALUM
ss -.00699
FYR
TEF .01450
k-ACT -.05018
d
HC INIT
N0x
7 ml -.52186
M $N02 .01009
EC/NO
X
kdADJ -.08556
CONST .05215
R2* 60
*Based on k,ACT and
d
Table 5-8 Reduced
2 3
74.5 78.6
-9.50805
-19.07977
10
-38
-1085
31
-212
-81
469
53
kjADJ,
.30616
.93828
-
-
.55029
.4699^
A3953
.42564
.45679
• 7
M
-365.99365
-
-
-
8.6o6o4
_
-302.70654
211.33298
79-5
the four material
Multiple Regression Models
6 7
0 MAX C OJTM
70.8 60.1
-.03980 34.26399
«.
.96338
-1.04420
-
6.08616
-.08095
_
.29652
3.80398
61.7
S/Vs, and % I
-196
-6l6
-921
-216
2967
54
*>2.
_
.48689
-
.49585
-
-
.18018
.76044
.64429
.9
8
0-DOSE
86.3
,.
373.43481
-313.43384
• -
2058.18018
-26.41008
_
189.46100
957.28271
67.4
9
HCFC
70.7
.07321
_
-.58105
-
-
n
HCT 50
75.4
-6.46031
14.12842
-99.00389
-
368.80396
-2.00685-2858.43530
.03853
_
-.37^57
.18939
70.4
47.56219
_
-115.42805
-468.91699
51
13
I/HE MAX R
67.2
-.00003
-.00596
_
-.08840
-
-
-.32312
.00587
_
-.02825
.06889
61.1
O
-------
LMSC-D406484
5.4 NORMALIZATION FOR k, EFFECTS
Q
The effect of variations in light intensity in smog chamber experiments is complex. It
is possible to use a photochemical smog simulation model and yary the light intensity.
One such simulation is reported by Niki (Ref. 7), for the propylene (2.23 ppm)/NO
(0.97 ppm) system at 90°F and 50% relative humidity. He found that the NO0 forma-
Z
tion rate and the time to NO0 maximum were nearly linear with light intensity, but
Z
that the run parameters that characterize later stages of the reaction are not. Thus
the propylene half time ratio was 1.67, and the ratio of ozone maximims is "much less
than a factor of two," for a reduction in k, of a factor of two. The data of Altshuler
(Ref. 5) show a ratio of ozone maximum concentration of about two, for a reduction
in light intensity of a factor of three in light intensity. It would not be surprising to
find an interaction of chamber material and size with light intensity effect.
One experiment was conducted to compare the effect of k, on the run parameters. The
empty baseline Pyrex chamber was used, and tests made at k, of 0.4 and 0.3 min
The data obtained is given in Table 5-9. If the parameter were linearly proportional
to k,, the ratio of the values would be 1.33 (or 1.33 for those parameters that de-
crease for an increase in k,). The difference between the experimental data and the
linearly proportional assumption is shown by the last column of Table 5-9, which gives
the ratio of the experimental data to 1.33. To account for the variation in k, among the
CAPI chamber, an adjustment factor defined as the product of the ratio of k, of the
chamber divided by 0.4 times the experimental data divided by 1.33 ratio was used
These run parameters are shown in Table 5-10, with the notation AA, DA, EA, etc.
A linear proportional adjustment of the run parameters is also shown in .the table.
These values are noted as AL, DL, EL, etc. These normalized data were used in a
stepwise regression of the CAPI data (as discussed below), but did not result in a
satisfactory fit of the data. It appears that k, effects cannot be accounted for by either
adjustment method. It would be interesting to use a propylene/NO photochemical
A
simulation model to predict the run parameters for a range of k, values from 0.16 to
-1
0.6 min , and to compare those results with experimental data.
5-13
LOCKHEED MISSILES & SPACE COMPANY. INC.
-------
IMSC/D406484
Table 5-9 Effect of k. Variation
a
Parameter
1. N02Rate
2. N02T MAX
3. N02DOSE
4. 0.,MAX Rate
5. 0~ Avg. Rate
6. 0- Max Cone.
7. 0 T Max
8. OJDose
9. HC Final Cone.
10. HC T.75
11. HC T.50
12. HC T.25
13. HC Max Rate
14. HC Avg. Rate
15. Aid Max
16. PAN Max
17. NO Rate
18. N02DF
19. 0 DF
20. FWHM
21. Cross Time
22. NOgXT
23. 0-XT
k,0.4
a
22.5
65
127
31.5
7.78
1.25
120
246
.10
52
82
in
25.4
18.1
1.15
.37
11.3
28.3
30.4
68
38
27
82
k 0.3 Ratio .4/.3 Exper Rat:
12.8
95
161
25.5
4.11
• 97
160
152
.16
88
129
. 162
20.3
11.2
• 75
• 38
7.44
36.9
3^.7
92
58
37
102
1.752
1.462'1
1.267'1
1.235
1.893
1.290
1.333"1
1.623
1.60"1
1.692'1
1.466'1
1.459"1
1.253
1.619
1.530
l.Ol"
1.512
1.267"1
1.125"1
1.353"1
1.526"1
1.370"1
1.240"1
1.31
1.10
.950
.926
1.420
.968
1.000
1.22
1.20
1.27
1.10
1.10
.94
1.21
1.15
•757
1.13
•930
.843
1.02
1.14
1.03
•930
5-14
LOCKHEED MISSILES & SPACE COMPANY. INC.
-------
Table 5-10 CAPI Chambers Adjusted to k of 0.4
1.
2.
53-
O
Z4.
m
05.
26.
87.
m 8.
(A
B>
AA
N02Rate 19
NOJTime 67
NOgDose 110
0-MR 34
o
0 AR 6
0_MC 1
3
0 TM 125
O^DOSE 214
3
AL
•9 19.5
67.5
109
.7 23.8
.69 6.49
.18 1.19
125
211
W9. HCFC .10 .10
TJ
o
°12.
|l3-
j>
z
n
HC25 71.6
HC50 95.5
HC75 129
HCMR 26.9
HCAR 15 .4
73.1
96.2
130
27.0
15.1
B
33
35
C
42
34
149 210
37
12
1.3
93
173
.01
39
54
78
4i
28
20
13
1.4
175
317
.02
29
55
76
33
29
DA
39.6
33.6
116
12.4
4.67
.44
86.1
53-7
.42
31.9
74.6
124
16.2
18.9
DL
40
33.6
116
12.4
4.67
.44
86.1
543
.42
31.5
74.6
124
16.2
19.0
EA
59-4
35
122
5.98
6.26
1.72
144
292
.26
19.2
39.2
60
28.9
31
EL
50
34
118
6.25
5.00
1.75
144
258
.26
22.4
41.6
63.6
30
27.5
FA
25
62
150
17.5
8.82
1.11
270
252
.008
46.9
80
120
21.8
22.1
FL
20
69.3
143
17.8
6.66
1.10
270
226
.01
54.9
81
122
23-3
20.0
GA GL HA HL IA IL JA JL
16.9 18.7 18.6 18.8 87.9 76 31-5 23.9
77-5 75 69.9 68.6 14.3 15 39-7 43-5
161 164 92.393.1 44.643.560.857.8
18.6 18.0 34.6 34.3 48.3 50 33-3 35-9
4.09 4.67 6.82 7.34 26.6 22 li.o 7.71
.67 .67 .98 .98 1.58 1.6 1.03 1.06
135 135 130 130 62 62 69.8 69.8
115 123 188 196 144 130 185 152
, .016 .015 .013 .012 .005 .005 .06 .08
78.5 72 56.6 53.9 14. 7 15.5 43.7 55-5
107 104 84.9 83.3 28.9 30 62.8 69
147 143 111 109 4i.9 44 77.5 84.8
21.1 20.7 28.9 28.6 50.4 52 47.6 50.5
13.1 14.0 17.3 18.0 55-3 50 25.9 21.3
01
t->
en
-------
LMSC-D406484
5. 5 COMPUTATIONS USING NORMALIZED CAPI DATA
An attempt was made to utilize the prediction formula coefficients given in Tables
B3-1 and B4-3 of Appendix B. These equations were obtained from the factorial
experiment, and apply to a k^ of 0.3 min~ , and to addition of each material at the
noted S/V to a base S/V of 1.4 ft"1 of Pyrex. Because of this structure of the
factorial experiment, it is not possible to use the prediction formulas directly in
analyzing the CAPI data. The factorial experiment chamber conditions that most
closely resemble each CAPI chamber was determined as found in Table 5-11. There
were 16 factorial experiment condition combinations to choose from (four materials
times two S/V values times two cleaning techniques) in matching each CAPI chamber.
The prediction equation was then used to predict the CAPI results which would be
expected after adjusting for the initial conditions of initial percent N(>2 in NOX and
hydrocarbon content.
Tables 5-12, 5-13 and 5-14 exhibit the predicted values, the residuals (or differences)
of the predicted values and the actual CAPI results after adjustment for k
-------
LMSC-D406484
The correlation coefficients between the run parameters and the ozone half lives are
given in Table 5-16. Ozone half-life data was not available for Chamber E. The
halflife for Chamber I seemed unusually long, and a correlation was performed that
omitted that chamber. The correlation coefficients obtained in the factorial experiment
are also shown. The differences show that ozone half-life by itself cannot be used to
characterize smog chambers.
5-17
LOCKHEED MISSILES & SPACE COMPANY. INC.
-------
LMSC -
Table 5-11
Material, Spectrum, S/V and Cleaning for CAPI Chambers
Chamber
A
B
C
D
E
F
G
H
I
J
Material
Pyrex
Aluminum
Teflon
Stainless Steel
Pyrex
Aluminum
Teflon
Aluminum
Stainless Steel
Stainless Steel
Spectrum
Full
Full
Full
Full
Full
Full
Full
Full
Full
Full
S/V
low
low
hi
hi
hi
low
low
low
hi
low
Cleaning
Purge
Purge
Vacuum
Vacuum
Vacuum
Purge
Purge
Purge
Vacuum
Purge
5-18
LOCKHEED MISSILES & SPACE COMPANY. INC.
-------
IMSC-D406484
Table 5-12
Predicted CAPI Run Parameters
Run Parameters
Chamber
A
B
C •
D
E
F
G
H
I
J
1/N02R
1
.0726
.O6o4
.1044
.0333
.0788
.O6o4
.0806
.O6o4
.0333
.0355
NOpTM
2
82.76
75.87
112.03
52.16
97.21
75.87
112.50
75.87
52.16
55.59
N02DOS
3
135-6
154.7
165.8
138.0
148.8
154.7
196.6
154.7
138.0
134.3
OoMAXC
3 6
.981
1.053
1.135
.890
1.003
1.053
1.101
1.053
.890
1.143
Table
Residuals from Predicted
OnTM
JT
123-5
.131.3
173-2
111.6
144.0
131.3
178.0
131.3
111.6
108.7
5-13
03IX)SE
8
165.6
183.8
17^.3
162.9
155-6
183.8
173.8
183.8
162.9
204.9
vs Normalized CAPI
HCFC
9
• 1519
.0705
.1085
.1095
.1159
.0705
.1289
.0705
.1095
• 0753
HCT50
11
132.1
95-9
163.8
75-8
89.6
95-9
105.4
95-9
75-8
75.2
1/HC MAX R
13
.0263
.0376
.0375
.0359
.0451
.0376
.0635
.0376
.0359
.0367
Parameters
Run Parameters
Chamber
A
B
C
D
E
F
G
H
I
J
I/NO R
1
.0160
.0352
.0508
.0068
.0528
.0036
.0138
.0109
.0473
.0055
N02TM
3-3
42.5
47.2
13.1
52.5
-7.6
16.5,
3-3
60.1
20.3
NOJX)S
3
-16.3
-53-2
-103-3
-22.3
-9.3
-40.8
-24.6
18.7
62.4
42.2
O..MAXC
36
.067
.039
.107
.565
-.3162
.2208
.5908
.2778
-.3680
.2860
O^TM
-32.5
21.0
-79-9
2.8
-47.4
-219.8
2.1
-22.7
57.4
38.9
0-DOSE
38
-1.3
41.3
-49.9
113.2
-39.2
8.5
78.3
21.0
63.0
70.5
HCFC
9
.0109
.0545
-.1075
-.4375
- .2281
.0435
• 1059
.0405
.0425
-.0437
HCT50
11
15.29
38.67
69.40
-17-0
34.8
-2.6
-28.9
5-6
86.5
8.19
1/HC MAX R
13
-.0227
.0046
-.0045
-.0491
.0001
-.0184
-.0005
-.0084
.0136
.0107
5-19
LOCKHEED MISSILES & SPACE COMPANY. INC.
-------
IMSC-D406484
Table 5-11*-
Normalized Residuals from Predicted vs Normalized CAPI Parameters
1/NDpR N02TM
Chamber 1 2
A
B
C
D
E
F
• G
H
I
J
Average of
Absolute .
Value
2.39
5-25
7-58
1.01
7.88
0.54
2.06
1.63
7.06
0.82
3-62
0.46
5.90
6.56
1.82
7.29
1.06
2.29
0.46
8.35
2.82
3-70
NOpDOS
3
-2.67
-8.72
-16.93
-3.66
-1.52
-6.69
-4.03
3.07
10.23
6.92
6.44
Run Parameter
OoMAXC 0-TM
°6 37
3-05
1.77
4.86
25.68
-14.37
10.04
26.85
12.63
-16.73
13.00
12.90
-4
2
-10
0
-6
-28
0
-2
7
4
6
.17
.70
.26
.36
.08
.22
.27
•91
.37
.99
•73
OoDOSE HCFC HCT50
°8 9 11
-0.21
6.56
-7.92
17-97
-6.22
1.35
12.43
3-33
10.00
11.19
7.72
0.31
1.56
-3.07
-12.50
-6.52
1.24
3.03
1.16
1.21
-1.25
3.18
2.11
5-33
9-56
—2 ^^
h YQ
-0.36
-3.98
0.77
11.91
1.13
4.23
Average
of
1/HCMAX2 Absolute
13 Value
-4.73
0.96
-0.94
-10.23
0.02
-3-83
-0.10
-1.75
2.83
2.23
2.76
2.23
4.31
7-52
8.40
6.08
5-93
6.12
3.08
8.4l
4.93
5-70
5-20
LOCKHEED MISSILES 8e SPACE COMPANY. INC.
-------
Table 5-15
CORRELATION COEFFICIENTS OF RUN PARAMETERS
1 2 36 7 8 9
NO2 Rate
0
O
^ J.
I .
m 2 .982 .933
O
2 3 .639 ,611
55
i2 6 -457 -033
r
S3 7 .654 ,852
™ vn
w V 8 .267 7652
TJ H
£ 9 -220 »672
o 11 ,938 ,922
O
| 13 .447 4608
NO2 T NO2 Dose O3 Max 03 T 03 Dose HC Final
Max Cone Max Cone
.588 .843
-447 -047
.679 .970
,233 -753
^196 .780
,929 .990
.495 ,785
j!93 T209
.650 .913
.319 -742
T191 ,755
.568 »852
.376 .871
-013 -147
.674 ,656
-313 -;394
-680 -.118
-782 -326
,555 -801
T208 .805 -439 -.811
.486 ,971 -037 -786 -;005 ,810
,317 .858 -456 -799 .590 .755
Z
Z
o
11
13
.657 .794
6
*>
o
05
ht"
oo
-------
Table 5-l6 Correlation Coefficients of Run Parameters and Ozone Half-Lives
CAPI CHAMBERS, EXCLUDING E
FACTORIAL EXPERIMENT DATA
5
n
i
m
m
D
2
55
0)
F
m
en
•o
O
m
0
o
2
TJ
z
z
o
1
2
3
6
7
8
9
11
13
l/NOpR
N02 TM
W02 DOS
03 MAX C
Do TM
Oo DOSE
HCFC
HCT 50
1/HC MAX R
Sample Size
cn
Light
•795
• 775
.411
.122
• 579
-.373
• 703
• 739
.296
Dark
.824
• 792
.480
.256
.628
-.226
.539
• 750
.300
8
8
1
2
3
6
7
8
9
11
13
NO_TM
NOZDOS
3
o:? TM
0^ DOSE
HCFC
HCT 50
1/HC MAX R
Samt>le Size
CAPI
Light
.583
.621
.151
• 359
..152
.003
•.519
-713
••315
9
ORIG
Dark
-.385
-.451
-.156
-.047
-.027
-.266
-.315
-.381
-.153
9
CAPI CHAMBERS
CAPI
Light
-.513
-.506
.51*
.782
-.164
• 533
-.549
-.741
-.317
ORIG
Dark
-.233
-.195
• 513
.830
.078
.864
-.517
-.4-03
-.123
CAPI
Light
-.555
-.5^3
.011
.706
-.295
.159
-.524
-.642
-.588
9
EXCLUDING
CAPI
Light
-.006
-.087
.656
.527
-.005
•372
-.577
-.235
-.427
ADJ
Dark
-.732
-.682
-.437
.701
-.361
.064
-.307
-.763
-.561
9
E AND I
ADJ
Dark
.241
.084
.445
.706
.188
.815
-.575
-.104
-.531
8
8
8
8
o
o
O5
*>.
GO
-------
LMSC-D406484
Section 6
RECOMMENDATIONS
Further investigations are recommended, as outlined below:
a. Conduct a similar set of tests for another hydrocarbon/NO system, such as
Ji
m-xylene/NO . This will indicate whether the observed spectral effects are
•X
general, as the range of organic species from unreactive aliphatic (butane),
reactive olefin (propylene), and reactive aromatic (m-xylene) will then be
available. Also use of a range of compounds from unreactive gas to polar high-
boiling liquid may show differing cleaning effects.
b. Perform further studies of the spectral effect, by varying the cutoff wavelength.
By use of Teflon rather than Pyrex chamber faces, the amount of light at wave-
lengths below 320 nm can be substantially increased. Evidently this lower wave-
length light is disproportionately important in smog photochemistry. A second
new spectrum to investigate is one that has a cutoff between the 320 and 350 nm
spectra just investigated. Having available spectral effect data for four cutoffs,
it is then possible to obtain an important function for the wavelength range.
c. Investigate light intensity effects by a set of tests at 67 percent and 200 percent
of the light intensity previously used. It is theoretically stated that initial
behavior of the photochemical system is linear with light intensity, but how late
smog manifestations, such as ozone maximum concentration and PAN build-up,
vary is not well known. Such data will also be helpful in applying chamber data
to the atmospheric diurnal intensity variation.
d. Searching for explanations of persistent anomalies in smog chamber behavior
would be productive. Among such anomalies not well understood at present are
the occurence of peak NO0 concentrations greater than initial NO charged (for
ft X
fast reacting systems such as propylene); the initial induction period in NO
disappearance; and the entire nitrogen balance. One technique for such an
investigation would be to utilize an alternative detection method for the nitrogen
species to correlate with the Saltzman NO_ and the chemiluminescent NO. Time
Derivate Spectroscopy is such a technique, and arrangements may be made for
such a spectrometer.
6-1
LOCKHEED MISSILES & SPACE COMPANY. INC.
-------
LMSC-D406484
e. Ozone production from low levels of hydrocarbons and nitrogen oxides is a sub-
ject of increasing practical interest, and could be investigated well using these
facilities. The effect of carbon monoxide in such a system is not well known.
"Background" runs with say 0.1 ppm NO and 100 ppm CO would be of great
interest.
6-2
LOCKHEED MISSILES & SPACE COMPANY. INC.
-------
LMSC-D406484
REFERENCES
1. Coordinating Research Council, Individual Hydrocarbon Reactivity Measurements:
State-of-the-Art, CRC Report No. 398, New York, Jun 1966
2. D. B. Wimmer, "Factors Affecting Reactions in Environmental Chambers,"
Coordinating Research Council Inc., Air Pollution Research Advisory Committee
Symposium, Chicago, May 1971
3. R. J. Jaffe, Factors Affecting Reactions in Environmental Chambers, Phase I,
LMSC-A997745, 20 May 1972
4. O. L. Davies (ed.), The Design and Analysis of Industrial Experiments, Ch. 7,
"Factorial Experiments," p. 247, Hafner Publishing Co., New York 1956
5. A. P. Altshuler and I. R. Cohen, "Structural Effects on the Rate of Nitrogen
Dioxide Formulation in the Photooxidation of Organic Compound-Nitric Oxide
Mixtures in Air," Int. J. Air Wat. Poll.. Vol. 7, 1963, p. 787
6. W. A. Glasson and C. S. Tuesday, "Hydrocarbon Reactivity and the Kinetics
of the Atmospheric Photooxidation of Nitric Oxide," J. Air Pollution Control
Assoc.. Vol. 20, 1970, p. 239
7. H. Niki, E. E. Daby, and B. Weinstock, "Mechanisms of Smog Reactions,"
Advan. Chem. , Vol. 13, 1972, p. 16
8. J. J. Bufalini, B. W. Gay, and K. L. Brubaker, "Hydrogen Peroxide Formation
From Formaldehyde Photo Oxidation and Its Presence in Urban Atmospheres,"
Env. Sci. Tech., Vol. 6, 1972, p. 816
9. P. A. Leighton, Photochemistry of Air Pollution, Academic Press, 1961
10. R. J. Gordon, in National Air Pollution Control Administration Pub. 999-AP-38
11. R. C. Hirt et al., Ultraviolet Spectral Energy Distributions of Natural Sunlight
and Accelerated Test Light Sources, J. Opt. Soc. Am., Vol. 50, 1960, p. 706
12. J. S. Nader, in National Air Pollution Control Administration Pub. 999-AP-38
13. M. W. Korth, A. H. Rose, and R. C. Stahman, "Effects of Hydrocarbon to Oxides
of Nitrogen Ratios on Irradiated Auto Exhaust," Part I, J. Air Pollution Control
Assoc., Vol. 14, 1964, p. 168
R-l
LOCKHEED MISSILES & SPACE COMPANY. INC.
-------
LMSC-D406484
14. R. J. Jaffe, et al., Factors Affecting Reactions in Environmental Chambers,
Phase m. LMSC-D401598. 28 April 1974
15. A. P. Altshuler and J. J. Bufalini, "Photochemical Aspects of Air Pollution, A
Review." Env. Sci. Tech. Vol. 5, No. 1, 1971, p. 39
16. A. P. Altshuler, etal., "Photochemical Reactivities of n-Butane and Other
Paraffinic Hydrocarbons," J. Air Pollution Control Assoc., Vol. 19, No. 10,
1969, p. 787
17. H. Jeffries, et al., "Photochemical Conversion of NO to NO2 by Hydrocarbons
in an Outdoor Chamber," Air Pollution Control Association, paper 75-16.2,
June 1975
18. M. C. Dodge and T. A. Hecht, "Ozone Decay in Irradiated Smog Chambers,"
submitted to Environmental Letters, June 1975
19. R. H. Sabersky, etal., "Concentration, Decay Rate, and Removal of Ozone,"
Env. Sci. Tech., Vol. 7, 1973, p. 347
R-2
LOCKHEED MISSILES & SPACE COMPANY. INC.
-------
LMSC-D406484
Appendix A
PHOTOCHEMICAL RUN DATA
A.I Propylene Graphs
Photochemical run data are given in this appendix in several forms - plots of NO, NO0,
&
ozone, and propylene vs irradiation time, and tabulations of the calculated run param-
eters as defined in Table 3-1. Data at 10 or 15 minute intervals are available for
acetaldehyde and PAN, but only the maximum for these species is reported in the
tables. The materials order is aluminum, Pyrex, Teflon, and stainless steel. Eight
run graphs are shown for each material. These are for the eight combinations of S/V,
spectrum, and cleaning investigated, and are arranged in sequence as follows:
S/V Spectrum Cleaning
High Full Vac
High Full Purge
Low Full Vac
Low Full Purge
High Cut Vac
High Cut Purge
Low Cut Vac
Low Cut Purge
For each material and variable combination, the average of the replicates of the indi-
vidual runs is given. Pages A-4 to A-ll give these run graphs.
A.2 Propylene Data Tabulation
Initial condition variations affect some of the run parameters rather strongly. Two
methods were used to account for initial conditions. One was based on photochemical
observations, and one on statistical methods. The photochemical observation method
accounted for the initial percentage of NO0 in the NO . As previously suggested,on both
& X
A-l
LOCKHEED MISSILES & SPACE COMPANY. INC.
-------
LMSC-D406484
theoretical and experimental grounds (Niki, Ref. 7, and B. Dimitriades, Bureau of
Mines RI 7433), this is accounted for by a linear extrapolation along the time axis
to the standard reference starting condition of 10 percent NO2 content in NO . The
tabulated run data show this adjustment as the column T ADJ. For the runs used in
-the effects analysis, the largest value of T ADJ is 14 minutes. Pages A-12 to A-17
give the data. The statistical analysis considered additional covariates and supported
this adjustment method. Appendix B discusses those analyses. Pages A-18 to A-22
give the covariate adjusted data.
It will be noted that two complete sets of experiments were performed for aluminum
surfaces. Changes were made in the instrumentation after runs 3 to 6 in the test
sequence. These changes made the set of runs between run 69 and run 85 better suited
for the effects analysis. In addition, as experience in operating the smog chamber
accumulated, better control of initial conditions were obtained. The data accumulated
in the later set of aluminum runs is preferable for the reasons just mentioned, and are
the ones used in the figures. Both the original runs and the reruns are tabulated for
the aluminum surfaces.
Run graphs showing the relative humidity effect are given on pages A-23 to A-30. Run
graphs for the butane runs are given on pages A-31 and A-32.
A-2
LOCKHEED MISSILES 8c SPACE COMPANY. INC.
-------
LMSC-D406484
This page is intentionally blank.
A-3
LOCKHEED MISSILES & SPACE COMPANY
-------
ALUMINUM
RUN S/V SPEC CLEAN
71 HIGH FULL VAC
ALUMINUM
RUNS S/V SPEC CLEAN
69.85 HIGH FULL PURGE
50
100 150 200
TIME (MINI
150 200
TIME (MINI
ALUMINUM
RUN S/V SPEC CLEAN
73 LOW FULL VAC
RUN
75
ALUMINUM
S/V SPEC
LOW FULL
CLEAN
PURGE"
50
100 150 200
TIME (MINI
250 300
100 150
TIME (WIN)
A-4
LOCKHEED MISSILES 8c SPACE COMPANY. INC.
-------
ALUMINUM
ALUMINUM
RUNS S/V SPEC CLEAN
3.0
S 2'°
Q.
Z
o
H-
<
oc
UJ
o
8 1.0
0
3.0
- 2.0
Q-
Z
O
t—
<
oe
1 1.0
0
t
81 HIGH CUT VAC
fc.
^
\
/
/
X
N
/
\
\
\
/-\
' \
^
/
NO
YLENE
" N02
OZONE
.v
K
\
>^^
^"*».
v_
^
~~-^_,
^--
) 50 100 150 200 250 3(
TIME (MINI
ALUMINUM
RUN S/V SPEC CLEAN
74 LOW CUT VAC
"\
v
X
"x
/
/
N
\
/
X
\
\
s\
N
/
/
PROPYLENE
NO
NO?
OZONE
\
V
^
\
*». .
\
^
3.0
i2'0
o_
z
o
i —
<
O£
UJ
O
8 i.o
• °o
3.0
CONCENTRATION (PPM)
.*** r°
'<=> 0
) 50 100 150 200 250 300
70,82,84 HIGH CUT PURGE
^
\
\.
/
/
\v
\
/
^
\
\
/^\
\
^X
/
/
V
\
X
/
x.
— — PROPYLENE
NO
N
0
^
°2
ZONE
--^
ll^x
50 100 ISO 200 250 301
TIME (MINI
ALUMINUM
RUN S/V SPEC CLEAN
x
\
\
/
72 LOW
N
/
y^
N
V
\
\
^
i
/
V /
A
CUT PURGE
PROPYLENE
NO
OZONE
V
\\
_ \
"*«»*
•**.
^>
*~-
'
50
100 150 200
TIME (MINI
250 300
A-5
LOCKHEED MISSILES & SPAGE COMPANY. INC.
-------
IMSC-1*06481*
PYREX
RUN S/V SPEC CLEAN
22 HIGH FULL VAC
PYREX
RUN S/V SPEC
21 HIGH FULL
CLEAN
PURGE
50 100 150 200 250 300
TIME (mini
100 150 200
TIME (mini
PYREX
RUN SIV_ SPEC CLEAN
31 •" LOW FULL" PURGE
PYREX
RUNS S/V SPEC CLEAN
27.33 LOW FULL VAC
100 150
TIME (MINI
100 150 200
TIME (mini
A-6
LOCKHEED MISSILES & SPACE COMPANY. INC.
-------
PYREX
RUN 5/V SPEC CUEAN
26 HIGH CUT VAC
CONCENTRATION (PPM)
«— fNJ V*» A
o . . o o o e
^
"'-•-.
^
\
/>
\
'/
(^
*m~r~-^^~ P
_.—. ,„ )y
0
\
/ N
V
ROPYLENE
0
°2
ZONE
\
/\
\
X;
) 50 100 150 200 250 30
TIME (mini
4.0
3.0
2.0
8
1.0
PYREX
RUN 5/V SPEC CLEAN
23 HIGH CUT PURGE
SO
PROPYLENE
NO
•NO,
OZONE
100 ISO 200
TIME (mini
230 300
PYREX
RUN SfV SPEC CHAN
30 LOW CUT VAC
PYREX
RUN S/V SPEC CLEAN
32 LOW CUT PURGE
50
100 150 200
TIME (mini
250 300
50 100 150 200 250 300
A-7
LOCKHEED MISSILES 8e SPACE 'COMPANY. INC.
-------
TEFLON
RUN S/V SPEC CLEAN
38 " HIGH FULL VAC
TEFLON
RUNS S/V SPEC CLEAN
34.35 HICH FULL PURGE
50 100 150 200 250 300
TIME (MINI
100 150 200
TIME (MINI
250 300
TEFLON
RUN S/V SPEC CLEAN
42 LOW FULL VAC
TEFLON
RUNS S/V SPEC CLEAN
43,44 LOW FULL PURGE
50
100 150 200
TIME (MINI
250 300
50
100 150 200 250 300
TIME (MINI
A-8
LOCKHEED MISSILES & SPACE COMPANY. INC.
-------
TEFLON
RUNS S/V
36,39 HIGH
CLEAN
"VAC"
50
100 130 200
TIME (MINI
300
3.5
3.0
z
o
<2.0
1.0
TEFLON
RUN S/V_ SPEC CLEAN
37 HIGH CUT PURGE
50
PROPYLENE
NO
•NO,
OZONE
100 150 200 250 300
TIME (MINI
3.5
3.0
2.0
1.0
. TEFLON
. RUN S/V SPEC CLEAN
40 " LOW CUT' VAC
X
X
I X
X
50 100 150 200 250 300
TIME (MINI
3.5
3.0
2.0
TEFLON
RUN S/V SPEC CLEAN
1.0
41 LOW CUT PURGE
\
' -s.
' >v.
^1
'X
/
\
/
/
X
'x
'x
PROPYLENE
_..^.__ k
0
X
N
/
/
X
X
X.
0
°2
ZONE
\
\
/
/
/
>-*—-
\
/\
/ \
0 50 100 150 200 250 300
TIME (MINI
A-9
LOCKHEED MISSILES & SPACE COMPANY. INC.
-------
STAINLESS STEEL
S/V SPEC CLEAN
HIGH
STAINLESS STEEL
RUNS SfV SPEC CLEAN
51,52.58 HIGH FULL PURGE
50
100 150 200
TIME (MIN)
50
100 150 200 250 300
TIME (MIN)
RUN
62
STAINLESS STEEL
S/V SPEC
LOW FULL
CLEAN
VAC
100 150
TIME (MIN)
STAINLESS STEEL
RUNS S/V SPEC CLEAN
61,66 LOW FULL PURGE
100 150 200
TIME (MIN)
250
300
A-10
LOCKHEED MISSILES & SPACE COMPANY. INC.
-------
STAINLESS STEEL
RUN S/V SPEC CLEAN
50
100 150 200 250 300
TIME (MINI
STAINLESS STEEL
RUNS SIV SPEC CLEAN
53,54 HIGH CUT PURGE
150
TIME (MINI
200
250 300
3.0
2.0
STAINLESS STEEL
RUN S/V SPEC CLEAN
~W~ LOW CUT VAC
8 1.0
\
PROPYLENE
----- NO
--- OZONE
50
100 150 200
TIME (MINI
250
300
STAINLESS STEEL
RUNS S/V SPEC CLEAN
60,67 LOW CUT PURGE
50
100 150 200
TIME (MINI
250 300
A-ll
LOCKHEED MISSILES & SPACE COMPANY. INC.
-------
LMSC-D406484
TITLE- ALUM
1*
2*
3*
4*
5*
6*
7*
8*
9*
10*
II*
12*
13*
14*
15*
16*
17*
18*
19*
20#
21*
22*
23*
RUNI
3
71
4
4
(S9
85
10
II
73
12
12
75
6
81
5
70
82
84
9
74
7
8
72
NO ST()\
.0
.0
.0
.5
.0
.0
.0
.0
.0
.0
.5
.0
.0
.0
.0
.0
.0
.0
.0
.0 -
.0
.0
.0
1 SPEC
—
—
— .
~
^
; CLEAN HCINIT
1 3.17
*
•»
^
™
—
—
3.10
3.05
2.90
3.04
2.94
2.96
3.03'
3.00
2.98
2.83
3.06
2.81
3.04
2.98
2.88
3. "07
2.96
2.94
3.02
2.84
2.92
3,00
NOX
.52
.52
.54
.41
.47
.49
.51
.43
.47
.51
.67
.45
.59
.49
.33
.47
.49
.41
.49
.46
.42
.36
.46
N02
.22
.25
.29
.29
.21
.13
.31
.22
.14
.11
.36
. 19
.31
.14
.12
.17
.17
.14
.33
.16
.22
.19
.15
PCN02
14. b
16.4
18.8
20.6
14.3
8.7
20.5
15.4
9.5
7.3
19.2
13.1
19.5
9.4
9.0
11.6
11.4
8.5
22.1
12.3
15.5
14.0
10.3
TAOJ
9
12
17
21
8
-2
20
II
0
-8
18
6
28
-2
0
4
4
0
35
6
16
12
0
HCNOX
2.09
2.00
1.98
2.06
2.07
1.97
1 .96
2.12
2.02
1.97
1.69
2.11
2.04
2.04
2.24
1 .96
2.06
2.10
1 .97
2.07
2.00
2.15
2.15
N02R
15. 10
14.00
15.80
13.00
14.70
16.00
15.80
13.20
13.80
13.10
lb.50
18. 10
10. 10
12.60
8.40
8.78
13.70
11.10
7.79
9.23
9.28
8.03
10.60
TfTtt
I
1*
2*
3*
4*
5*
6*
7*
8*
"0*""
10*
Jl#
12*
13*
14*
15*
16*
17*
18*
19*
20*
21*
22*
23*
.^>k
/T\
s~\
_/^\
sr\ .
.s*\ „__
.XT\
x~\
r-'"A'L'0'M""\Z; (& ^ ^ <^ ~"\ZJ <£T" W~~~ ^
1UN2NO N02TM N02DOS 03MAXR 03AVGR 03MAXC 03TM 03DOSE HCFC HCT75
3.0
71.0
4!5
"6970"'
85.0
10.0
11.0
"7370"
12.0
12.5
75.0
6.0
81.0
""575'"
70.0
84!o
9.0
74.0
7.0
8.0
" 72.0
79
94
'"77"'
81
""83"'
88
81
86
""9'l""
92
92
76
125
118
"107"'
144
"1T4"'
120
145
136
124
125
120
142
165
128
166
145
146
"T60"
159
168
157
178
211
123
200
'"2'06"
186
176
188
T5'9
140
T 1 72
22.8
21.3
"3T.'5"'
30.9
•"2T.'9"'
22.1
30.3
37.3
35.6
31.7
28.9
34.7
19.2
17.4
17.6
18.0
\3\B
23.0
26.2
22.8
26.7
21.2
4.46
4.55
5!53
"4748"'
4.23
6.09
5.54
5.18
5.38
5.31
5.47
3.24
3.29
2.72
2.71
"3702"'
2.73
2.74
3.48
3.32
3.38
3.52
.82
1 .00
"1704"
1.05
^93
.15
.12
..........
.14
.04
.04
.92
.94
'"777"
.90
'"784"
.83
.92
1.10
.91
.96
1 .02
129
152
T2'2""
121
143'"
14R
140
131
137
138
131
190
198
"T65"
219
•1'8'9""
210-
205
206
176
177
190
135
165
"165"
158
144
158
200
190
"l'8l"
189
193
201
133
140
114
112
— T'53"
121
116
147
127
135
137
.05
.18
.10
.10
".14"
.10
.19
.15
.12
.13
.14
.10
.15
.18
.35
.31
".'2"2'"
.27
.31
.16
.27
.28
.17
70
86
69
64
""76'"
82
81
83
91
85
91
63
101
104
•"96"
107
'TOT"
88
148
111
128
132
116
A-12
LOCKHEED MISSILES & SPACE COMPANY. INC.
-------
LMSC-D406484
TITLE- ALUM Qj)
1*
2*
3*
4*
5*
6*
7*
8*
9*
10*
II*
13*
14*
15*
16*
17*
18*
19*
20*
21*
22*
23*
RUN3NU
3.0
71 .0
4.0
— -475-
69.0
85.0
10.0
1 1 .0
73.0
... T--;0-
12.5
""7570"
6.0
81.0
5.0
70.0
82.0
84.0
9.0
74.0
7.0
870"
72.0
HCT50
100
122
99
— T0"5
108
— ll~8
109
1 1 1
120
118
148
153
137
165
144
-—[4-7
180
161
159
151
Q2) Q3) (M
•HCT25"HeMA50THCAV
127 27.4 15.
152
131
""T32""
143
149
141
143
146
— ~|-42—
145
—--,-24"
184
185
190
— 2IT"
186
— -,-94—
210
195
204
~~" 2~03~"
182
24.
25.
23.
22.
22.
24.
27.
25.
27.
26.
27.
18.
17.
17.
13.
18.
14.
23.
17.
21.
21.
23.
9 12.
1 15.
o 13!
9""["27
8 13.
2 13.
7 12.
0""T27
3 11.
3 ^ to!
0 9.
8 10.
2""¥:
2 10.
8 9.
6 7.
9 d.
0 8.
7 lo!
)
Qs)
Q$)
($
G"R" ALHM AX -"PANM"A"X"N()H"ATE '
80 .60 .56 12.50
30
40
70
60
50
20
JO
20
80
50
60
30
61/
70
10
10
89
54
37
63
10
.70
.55
""."51"
.72
.83
.59
.60
1.00
.~5"0
.54
.86
.77
.86
.94
.92
.82
.49
.90
.52
!98
.31
1.10
-T;TO"
.38
.29
.13
.17
.35
"".30"
.33
!23
.35
.32
.37
.33
.26
.18
.45
.16
""".T4"
.50
10. 1O
14.20
12.20
10.50
8.70
12.80
11.10
9.56
II .00
11.80
TO. 80
9.93
7.88
9.03
6.70
/ . 70
6.90
7.36
/.04
8.77
•"7782"
7.62
Q8)
N02W
31.1
35.5
31.6
30.3
33.6
35.9
32.1
34.0
35.5
35.1
33.5
34.7
37.3
46.0
30.9
45.6
45.2
"4371"
39.4
42.1
37.3
"3473"
38.5
()'3DF
28.5
30.5
36.7
•3-7.3
31.8
34 .2
44.2
44.3
40.2
41 .7
38.5
44.4
27.9
30.5
28. .5
25.6
29.2
28.0
26.0
32.9
29.8
33. f
30.6
•TTT1
2*
3*
4*
5*
6*
8*
9*
10*
II*
12*
13*
14*
15*
16*
17*
18*
19*
20*
21*
22*
23*
:E-"AL"Crv
RUN4NO
T.'O"
71.0
4.0
4.5
85!o
""1070"
11.0
73.0
12.0
12.5
75.0
6.0
81.0
5.0
70.0
""8270"
84.0
74 !o
1.0
8.0
""7270"
.
FWHM
86
66
70
— 89"
85
-—7-4-
67
82
72
73
77
no
115
89
126
""HO"
131
"117"
117
101
97
"T09"
/£75\
—®"—\&-
XTIME N02XT
43
52
44
47
—-44—-
49
"""43""
49
52
52
48
40
66
58
64
76
-— 5A— -
63
73
64
71
"—(53""""
36
42
33
34
39
39
.38
37
40
44
36
59
60
68
58
57
63
63
(SO
54
5/
03XT
100
78
74
99
-—9-7—
82
91
85
-"90—
91
124
140
101
143
""138""
147
"T23""
133
112
106
~1?7"~
-- :
A-13
LOCKHEED MISSILES & SPACE COMPANY. INC.
-------
LMSC-D406484
TITLE- PYREX
KiliJJJJO STOV SPEC Cl HAH HCTMTT
\#
4#
5* -
22
?1
27
.33.-
31
26..
25
1
1
-1
-1
-1
L
1
-1
1
1
1
1
1
-1
-1
-1
1
-1
1
J
-1
1
-1
1
3.22
2.94
3..QQ
2.84
2.9J3
2.95
JJUJC
.68
.55
-52
.39
..44
.54
JJU2
.27
. 1 7
.18
.16
.14
.18
.21
. 1 6
HCJ102.IADJ..HC1UUX
16.1 12 1.92
10.R 0 ?.04
11.6
1O.5.
10. 1
J2.S.
13.6
10.7
.. 0.
0
7
10
1 .90
1.9.7
2.04
2.07.
1 .92
?. 1 ^
(i)
N02R
12.80
1 1.40
13.50
15.00
10. 10
6.25
7.26
7.4R
32
-I -I
-1
3.00 1.43 .20 14.0
2.10 8.10
TITI H- HYRFX^' (£) \±) \2J (6J (7} W—
RUN2NO N02TM H02DOS 03MAXR 03AVGR 03MAXC 03TMAX 03DOSH
-1£ 22 99. 163—-2a.B—4.19 .98 15.2—152.9-
2# 21 95 161 24.7 3.85 .84 143 127.5
3*. 27. .78. LIA —-M.3— _d.b\6 L.16 122...187..7.
4# 33 79
<^4t 11 10?
114
1??
32
.9
6
•
13
1.11
.QR
125
140
190
.0 .11
81
R7
6# 26 192 225 17.7 2.17 .92 267 80.7
-7#- .25—-212 332—10..6— .U4.2 .73 310—.-4.U5
8# 30 147 161 28.5 3.02 .97 207 127.8
.9#_ 32—.121 14_4_-.-2S.A_—a.3.6 .9.1 Jj56-12A.b-
HCFC h
--16-
.26
.OB
. 11
.30
..-6-4...
.26
.--.24.
1CT75
...98
89
. 7Q
81
R7
149
-185-
133
..111
TITLE-
?#
22 130
159
26.8 12.00
.70
.15 10.30
•....03DJ-:
32.0 30.3
1
3#
.At...
5#
-6J— .
7#
27
—,33.
31
25
88
98
120
21D
249
17?
117
128
150
235
295
34.0
34-6
36.8
12.. Q.
12.1
16.40
J.5. LO
11.70
6^.62
5.20
R.9?
.96
.77
.68
.11
.13
.J4
.18
.T2.
12.70
14.20
9.19
5.Q7
5.00
7. 10
24.8
24.8.
29.1
52. L
72.0
43.4
41 .1
37.4
9.0
?R.4
32 148 183 20.6 9.52
.65
12
7.81 33.6 29.5
TITTF- PYRKXUg) (2J)_
RUN4NO FWHM XTIME
-L#. 22—.7-4 61
N02XT 03XT
-42—-91.
2# 21
.3* 27.
4# 33
c;rf II
77
-A3_-
66
77
55
-50_
50
40 88
-28.—72.
29 75
4? Rn
6# 26 138
_7& _______ 25— -176—
8# 30 115
_9i _______ 32—100
106 86 161
OD— _1L4— 2ia.
89 59 118
73 ______ _4R— .93.
A-14
LOCKHEED MISSILES & SPACE COMPANY. INC.
-------
LMSC-D406484
TITLE- TEFLON
RUM I NO STO\
1* 38
2# 34
3# 3b
4# 42
5# 43
6#
7#
8#
y#
I0#
II*
44
3ft
39
37
40
41
/ SPEC
_
:
-
: CLEAf
4 HCIMIT
3.00
2.82
2.86
2.97
2.98
3.03 '
2.99
. 2.-9(T
2.86
2.96
2.96
NOX
.52
.38
.50
.54
.52
.47
.63
.45
.48
.49
.47
N02
.16
.13
.15
.18
.15
.15
.17
.17
.15
. 17
. 14
PCN02
10.5
9.4
14. /
11.7
14.5
13.6
10.4
11.7
10. 1
1 1 .4
9.5
TAOJ
0
-2
14
5
13
10
0
7
0
5
-2
HC1NOX
1.97
2.04
1.91
1.93
1.96
2.06
1 .84
2.04
1.93
1 .99
2.01
NO
8.
II.
II.
9.
II.
10.
6.
5.
8.
6.
7.
&
72
30
10
52
60
50
19
91
II
8/
56
IIILE- iEFi.()N(g) y; (A) O (£) (j) (s) Cs) 35
4# 42
"7*
I0#
43
44
36
39
37
40
121
102
104
142
1 16
113
T8"0
203
762 ""
193
176
151
f6"7 "
205
203
188
228
223
267
22.5
23.6
"2"4."8"
31.3
32.9
26.2
"2"0:2~"
19.6
"T5.~7"
16.4
4.04
4.33
4.09
4.08
4.08
3.90
"2742""
2.25
""2728""
2.33
1 . 10
1.04
"!~.~0~4"
1 .25
1.12
1 .10
"1VOT"'
.98
--.85"
1.05
183
168
174
200
176
195
267
—•237""
295
166.0
166.0
181.0
189.0
189.0
180.0
94.8
97.0
97.9
93.0
.12
.09
.12
.22
.08
!32
.32
.30
.34
109
95
102
1 12
105
98
136
166 '
"139
147
41
205
303
10.8
1 .67
T8CT
298 5070.56
136
TITLE- TEFM)N
0
RUN3NO HCT50 HCT25 HCMAXR HCA
l# 38 143 174 23.6 10
2# 34
3# 35
4# 42
5# 43
6# 44
7# 36
8# 39
9# 37
10# 40
11* 41
1 33
134
T56
143
144
205
224
197
217
230
169
190
175
177
246
262
241 ,
263
282
21.3
21.7
lb.8
13.4
18.7
12.7
14.0
13.1
12.0
8.5
10
10
9
10
10
6
6
6
6
5
VGR ALDMAX PANMAX NORATE
.20 .70 .31 8.61
.50
.30
.05
.30
.20
.81
.1 1
.70"
.39
.80
.74
.95
..82
.79
.73
.70
.72
.72
.73
.72
.25
.23
" "."6T"
.39
.34
.25
.21
.18
.25
.12
8.86
8.77
1.99
8.67
7.86
6.24
5.33
6.65
5.92
5.41
N02DF
38.6
36
36
44
44
42
44
52
50
59
68
.4
.5
.3
.3
.4
.4
.9
.1
.8
.6
03DF
36.4
40.6
39.5
39Y8"
39.6
39.8
19.4
21.0
22.0
19.8
11.3
TITLE2 TEFLON
RUN4NO FrtHM
2!
4#
6*
8#
?0#
lit/
38
34
35
42
•"43
44
36
39
— --j?--
40
41
90
86
91
84
102
143
142
T40""
165
190
78
58
"""63""
73
" "63 '
60
107
107
96
- -94-
43
42
41
70
"53"
53
85
85
"5"8"
97
"III
105
no
..-m
127
"122 ""
122
157
157
"T55 •."" " "
199
2"0"4
A-15
LOCKHEED MISSILES & SPACE COMPANY. INC.
-------
LMSC-D406484
TITI h-
Itf
2#
3#
4#
5#
6#
7#
8#
9#
I0#
ll#
I2#
13*
I4*»
RUN 1 NO STO\
55
56
51
52
Sfl
62
61 ... T.
66
57
53
*>4
63
. .6.0.. ...-
67
/ SPEC
__.
-.
: CLEAf
-
..
J HCINIT
3'. 1 9.
3.04
. 1*OSL-
3. II
P.RA
3.02
— . 2*9fl
3.06
. . -3^11.
3.05
1.P4
2.88
. -3*02-.
3.00
NOX
.48
.52
*42.
.43
.40
.46
*_5Q-
.48
,44..
.49
-S?
.44
L*35.
.53
N02
*15
.IB
-15
.15
. 17
.15
. 16
.16
^17_
.16
. 17
.13
*ia
.18
PCN02
10.1
1 1.8
— 1Q..6
10.5
1 1 .4
10.3
...10*7
10.8
. U..8
10.7
11.?
9.0
- -IU6.
11.8
TADJ
0
0
0
0
n
0
Q
0
0
0
?
0
.3
3
HCNOX
2.16
2.00
2..1E
2.17
1 .9?
2.07
1.59 .
2.07
2.16
2.05
?. l ^
2.00
.. 1*25..
1 .96
N02R
33.5
29.1
49.2 .
45.7
41 - 1
17.8
.26*8
24.4
21.0 .
35.3
?rt-4
10.8
21*2..
19.2
TITLE-
Rl
3#
.5*. ...
I0#
I3#
144*
-. .SSTEEL©
JN2NU N02TM N
55 51
56 53
51 38
52 37
...58 4J_
62
6.1
66
57
53
54
63
60
67
73
58
63
86
53
65
1 11
81
80
®
fa
/AY -.
fT\
02DOS 03MAXR 03A^GR 03MA^C ()3TJfAX ()3D&§E H(
-133_. 23*2. ... 6*77. .*8& 115 16_2___
134
121
116
-.133 ...
139
135
134
IR6
157
164
181
174
158
20.9
22.2
21.6
._ J.9.*5_ _
31.1
3U7
12.5
12.6
20.0
16^5
19.6
6.87
8.16
8.04
.. 7*50..
6.86
- 8,.QO
7.70
3.43
4.72
- 4,40
3.50
4,34
4.50
.91
.81
.82
....-*85...-.
1.18
1.15
.69
.68
.96
,85
.89
110
90
95
.-UD..
130
_ LJQ
120
163
135
137
190
143
169
159
146
16.4-..
200
194
210
1 18
130
133. _
136
147
147
:Fc HC
10.
II
10
15
-14—
10
08
12
14
17
2Q
27
LI
27
- Qf
T75
.51
49
44
39
..44.
71
50
58
71
48
64
1 M
61
.64
.TITLE-.. SSTti
RUN3NO h
l# 55
3#
4# .
6#
7#
51
I0#
I2#
56
52
62
61
66
57
53
54
63
60
:L O
12
11
t
..A3.....
___1<3—
iCT50 HCT25 HCMAXR HCAVGR ALDMAX PANMAX
73 1O3 33.0 PI. 5 .60 .?3
77
7L
64
71
94
79
82
112
86
9-7-
148
1 11
107
JQ4
100
J04
124
1 10
113
L55
132
J46
190
146
26.7
2A.3
26.9
2S..3.
29.8
25.4
28.8
J.9.,.9.
18.5
20..5-
18.7
19.
22,
23.
19,
15.
18.
18.
13.*
17.
-J6,
9.
1 1.
3
1
5
7
7
R
2
7
6
,1
2
.74
.85
.91.
.82
.79
.82
.Bl ,
.79
....-ll_-_
.61
.79
.27
,23
.25
.21
. IR
.26
—,20
.16
.29
— _o~..
NORATE
17.90
18.10
Jtfj-80
20.00
19,40.
14.30
16.80
15.00
16.00
9.03
i ?. no
JJ
N02DF
30.0
29.4
28 ..3
27.1
.29*8. ..
31.8
30.2
-43*0...
35.0
_J5*9-.
41 .9
37.. 1
03DF
36.4
32.0
39 .V
45.Y
43.0
47.3
27,-4.
29.1
29*.l-
31 .6
31 .7
67
113 148 18.5 12.9
.89
.28 11.00 33.7 31.5
A-16
LOCKHEED MISSILES & SPACE COMPANY, INC.
-------
LMSC-D406484
TITIK- SSTm t>0 CD P2 G\
RUN4NO
U 55
2# 56
3# 5 1
4* 52
6*
9#
1 1 *
I4#
62
61
66
57
53
63
60
67
FrtRrrx
70
67
60
61
78
7b
69
107
88
HO
1 13
100
97
TIME N02XT 03)?t
25 26 .90
26 27 ' 84
17 21 78
17 20 78
18 5t> °5
38
27
31
.40
23
2°
59
38.
40
35
31
32
46
30
43
40
92
83
89
123
112
131
J.15 -
103
A-17
LOCKHEED MISSILES 8t SPACE COMPANY. INC.
-------
Covariate Adjusted Data
ALUMINUM
HUM NO
71
69
"5
73
75
81
T0
8?
84
74
77
3'JN NO
71
69
85
73
75
81
70
8?
84
74
72
RUN Nil
71
69
85
73
75
81
70
82
84
74
72
"UN NO
71
69
85
73
75
81
70
82
84
74
72
S/V
2
7
2
1
1
2
2
2
2
1
1
NH7TM
93.
80.
80.
85.
72.
114.
136.
109.
110.
130.
115.
HCT50
121.
108.
110.
114.
89.
148.
159.
139.
137.
155.
146.
SPrC CLCA\
7 '
•» I
2 I
•> 2
7 \
1 7
I 1
1 1
1 I
1 7
1 1
ALUMINUM
ND700S :
163.
152.
171.
164.
154.
215.
201.
709.
197.
187.
175.
ALUMINUM
HCT25 • '
151.
140.
141.
140.
120.
180.
203.
180.
184.
189.
177.
«LUMINUM
1 H
)3MA
20.
?1.
23.
37.
33.
17.
18.
77.
14.
76.
21.
^MA
23.
21.
23.
75.
26.
16.
13.
17.
15.
17.
23.
3.10
1.04
2.94
3. TO
1.06
3.04
7.38
1.07
7.96
3.0?
3.00
yi.
74
17
15
17
78
64
45
03
16
11
55
XR
53
51
70
91
53
86
61
83
04
76
81
CHHM XTIME N07XT
85.
89.
86.
83.
77.
116.
126.
110.
132.
117.
110.
50.
41.
42.
46.
36.
53.
69.
50.
53.
67.
58.
42.
39.
39.
40.
36.
61.
67.
59.
57.
63.
57.
r NOX
1.5?
1.4?
1.49
1.47
1.45
1.49
1.47
1.49
1.41
1.46
1.46
T3AV»nB
4.75
4.69
4.46
5.40
5.1?
3.44
'.78
3.7?
2.8?
3.38
3.68
HCAVGR
12.29
17.32
13.12
12.3?
14.73
10.15
8.61
10.37
10.62
9.06
10.44
03XT
100.
99.
99.
98.
92.
141.
147.
139.
147.
133.
177.
MO?
.75
.'1
.13
.14
.19
.14
.17
.-17
.14
.1ft
.15
TN07 TA
16
14
8
9
13
9
11
11
8
17
10
03MAXC
I.
0.
0.
1.
1.
0.
0.
0.
0.
1.
1.
00
94
94
14
05
96
39
86
84
10
03
ALDMAX
•
•
•
1.
•
•
•
•
•
•
•
73
73
81
00
83
88
90
95
81
91
98
nj HC/MOX MH'0
.4 12. 2.
.3
.7
.5
.1
.4
.6
.4
.5
.1
.3
03TM4X
151.
140.
140.
143.
127.
194.
211.
184.
200.
200.
185.
PANMAX
.29
.37
.37
.38
.70
.39
.36
.35
.29
.46
.52
8. 7.
2. 1.
0. 7.
6. 7.
2. 7.
4. 1.
4. 7.
0. 7.
6. •>.
0. 7.
oi^os11
148.9
150.1
160.3
182.6
1°0.6
142.1
107.2
179.9
125.0
137.6
116.1
9.4t
10.18
S.93
10.73
11.27
8.75
7.12
8.17
7.4?
7.11
7.95
00
07
9?
0?
11
04
96
06
10
1?
l"
Hcpr
.16
.13
.11
.12
.19
.18
.33
.71
.28
.16
.17
; Nn?n«
35.9
33.9
36.7
35.9
34.7
46.4
45.5
45.6
43.4
41.8
38.7
13.50
15.31
18. 3«
14.68
18. IP
13.19
9.40
14. "4
I?.1??
9.10
11.11
HCT75
87.
77.
76.
86.
58.
go.
101.
96.
80.
105.
111.
: 030*
3?. 5
33.4
33.8
39.9
42.1
30.7
24.6
78.1
28.1
30.8
'0.0
A-18
LOCKHEED MISSILES & SPACE COMPANY. INC.
-------
Covariate Adjusted Data
R'JV NfJ
22
21
27
33
31
?6
25
30
32
S/V SPP.C ntr«M HCTNTT NOX
2
2
1
I
1
2
2..
I
1
2
2
2
2
2
I
I
1
I
2
1
2
2
1
7
1
2
1
3.?2
3.20
2.94
1.10
2. 34
2.98
2.05
3.23
3.00
1.68
1.57
1.55
1.52
1.39
1.44
1.54
I. 50
1.43
NH2
.27
.17
.18
.16
.14
.18
.21
.1.6
.20
*N02 TAOJ H^/NOX N0'9
16
10
11
10
10
12
13
10
14
.1 12
,P 0
.6 5
.5 0
.1 0
.5 7
.6 10
.7 2
.0 11
I.
2.
1.
I.
• L •
2.
1.
• ? •
2.
o?
04
?0
97
04
0T
92
!c
10
11.67
13.55
1 6.7<;
If-. 13
11.06
6.31
7.54
7.70
8.20
OYRPX
*JN NO
72
71
27
33
31
26
25
30
32
MD7TM
103.
95.
70.
75.
94.
135.
206.
145.
113.
^32^09
157.
164.
117.
117.
125.
225.
330.
164.
142.
T3MA
7ft.
23.
37.
33.
30.
17.
10.
27.
25.
XR
?5
98
59
78
06
76
60
40
19
D3AVGR
4.57
4.27
7.03
6.«?7
4.63
2.23
1.52
3.28
3.63
I.
0.
1.
I.
0.
0.
0.
1.
0.
00
qp
16
12
Q7
92
72
01
90
03TM4X
152.
143.
114.
120.
137.
260.
302.
205.
158.
03?HSC
153.3
133.1
198.o
193.7
158.5
80.4
44.4
134.7
130.?
HCCC
.13
.2*
.09
.11
.1.5
.30
.64
.23
.24
HCT75
93.
95.
63.
77.
82.
142.
178.
127.
103.
PY5PX
*JN NO
22
21
27
33
31
26
25
30
32
P'
RUN NO
22
21
27
33
31
26
25
30
32
HCT50
131.
124.
80.
93.
ill.
203.
241.
170.
141.
PWHH
73.
78.
63.
66.
77.
138.
176.
116.
100.
HCT25
159.
155.
109.
123.
142.
248.
287.
199.
176.
XTTMP
58.
52.
43.
45.
54.
99.
93.
85.
66.
HCMAXR
23.
21.
35.
34.
41.
12.
12.
20.
20.
92
51
46
84
67
05
22
92
53
N02XT
1
43.
42.
28.
29.
41.
86.
13.
61.
48.
HC4VGR
11.59
11.00
18.28
15.80
12.84
6.93
5.43
8.77
10.06
03XT
93.
90.
72*
75.
79.
161.
209.
121.
93.
ALPMAX
.
•.
.
.
.
.
.
.
•
78
75
94
88
72
63
66
65
65
PANMAX
.15
.20
.11
.13
.13
.14
.16
.11
.11
NORAT
10.03
10.09
14.16
15.29
9.77
5.25
5.05
7.41
8.1&
N02DF
32.0
34.7
25.0
25.0
29.1
52.1
71.5
36.4
33.5
P30F
3^.3
27.4
42.8
41.5
37.9
18.7
11.0
28.7
29.9
A-19
LOCKHEED MISSILES & SPACE COMPANY. INC.
-------
IMSC -
Covariate Adjusted Data
TFFLON
?IM NT
38
34
^5
42
43
44
36
39
37
40
41
S/V S°PC CIF«N ' •HriNtf NOX
2
2
2
1
1
1
2
2
2
I
I
2
7
?
•>
2
7
1
1
1
I
I
2 3.00
1 7.9?
1 2.86
2 2.97
1 2.98
1 3.03
2 2.99
2 7.96
I 7. 86
2 2.96
1 2.96
1.
1 .
1.
I.
I.
1.
I.
1.
I.
1.
1.
52
39
50
54
52
47
63
45
48
49
47
N02
.16
.13
.15
.18
.15
.15
.17
.17
.15
.17
.14
TNO? TfiOJ Hf/NOX
10
q
14
11
14
13
10
11
10
11
9
.5 0
.4 -2
.7 14
.7 5
.5 13
.6 10
.4 0
.7 7
.1 0
.4 r>
.5 -•>
1.
7.
1.
I.
1.
2.
I.
• 7 •
1.
1.
• • . •
97
04
91
9'
96
06
94
04
93
90
01
NO?R
9.00
12.45
17.7*
10.25
12.99
11. 3P
6.39
6. 20
0.69
7.27
7.84
TFFLDN
'UN NO
38
34
35
42
43
44
36
39
37
40
41
T
3IJN NO
39
34
35
42
43
44
36
39
37
40
41
HJN NO
38
34
35
42
43
44
36
39
37
40
41
M02TM
117.
92.
92.
136.
107.
111.
175.
194.
154.
195.
200.
-FLON
HCT50
138.
124.
123.
149.
135.
137.
203.
214.
189.
209.
224.
CF=MHM
90.
87.
90.
84.
93.
102.
143.
142.
140.
165.
191.
N0200S
170.
155.
162.
?06.
199.
195.
220.
229.
226.
269.
307.
HCT25
169.
153.
159.
183.
167.
170.
241.
253.
233.
255.
276.
XTIMC
73.
52.
54.
66.
55.
58.
102.
99.
76.
88.
89.
C13MAXR
72.94
25.25
75.06
31.95
32.26
75.71
70. 5P
19.99
16.76
16.61
10.98
HCMAXR
23.^0
23.09
'2.8P
19.01
13.42
1 8. 45
12.77
14.16
13.61
12.12
8.59
N02XT
43.
41.
40.
70.
52.
53.
85.
85.
57.
97.
111.
H3AVGR
4
4
4
4
4
4
2
2
2
2
1
.23
.35
.18
.26
.28
.13
.49
.3?
.30
.41
.70
03M&XC
I.
I.
1.
1.
I.
1.
1.
0.
0.
I.
0.
11
03
01
25
11
10
02
98
P4
05
81
HCAVGR ALOMAX
10
11
11
9
10
10
6
6
7
6
5
.52
.38
.53
.51
.99
.71
.97
.41
.03
.66
.93
03XT
105.
109.
109.
127.
121.
122.
157.
157.
154.
199.
204.
.
.
.
.
.
.
.
.
.
.
.
70
68
90
91
78
74
70
71
67
72
71
03TM/VX
179.
160.
163.
193.
167.
188.
255.
758.
279.
287.
293.
PANMAX
.33
.25
.19
.61
.37
.33
.27
.21
.18
.26
.14
0300SE
169.7
169.6
1 65.7
185.1
177.0
170. '
99.6
92.2
100.6
90.1
54.8
NORATF
«.00
9.35
9.50
8.43
8.74
8.15
6.45
5.58
6.94
6.10
5.54
HfF'
.12
.12
.13
.73
.08
.14
.32
.33
.32
.35
.57
38.8
'6.4
'6.4
44.3
44.1
42.2
44.6
52.9
50.1
59.9
68.8
^rT75
105.
91.
93.
105.
96.
90.
132.
157.
134.
140.
132.
r 030F
3*.. 8
40.7
37.5
40.0
39.2
38.7
19.8
21.7
22.5
20.0
11.9
A-20
LOCKHEED MISSILES & SPACE COMPANY. INC.
-------
Covariate Adjusted Data
LMSC-D406484
STftlNLFSS STEFL
*UN NO
55
56
51
52
58
62
61
66
57
53
54
63
60
67
S'
tUN NO
55
56
51
52
58
62
61
66
57
53
54
63
60
67
S/V SPFC CLFftN HCINTT NOX
2
2
2
2
2
I
1
1
2
2
2
1
1
I
S02TM N
49.
5'.
35.
35.
38.
69.
54.
60.
93.
50.
65.
101.
76.
75.
STAINLESS
*')N NO
55
56
51
52
58
62
61
66
57
53
54
63
60
67
HCT50
70.
76.
68.
61.
66.
99.
74.
79.
109.
82.
96.
137.
106.
108.
7
2
2
2
2 ,
2
?
•>
1
1
1
1
1
1
STFFL
2
2
1
1
1
2
I
1
2
1
1
2
1
1
3.19
3.04
3.09
3.11
2.86
3.02
2.98
3.06
3.11
3.05
3.24
2.88
3.02
3.00
1.48
1.52
1.42
1.43
1.49
1.46
1.50
1.48
1.44
1.49
1.52
1.44
1.55
1.5?
02005 H3M&XP fl3ftVG"»
137.
135.
124.
119.
134.
142.
137.
137.
187.
160.
166.
186.
175.
159.
STFFL
HCT25
100.
106.
101.
97.
100.
119.
105.
110.
152.
128.
144.
IPO.
141.
143.
22.
20.
??.
21.
20.
31.
33.
31.
12.
12.
13.
21.
16.
19.
HCM&
29.
26.
25.
25.
30.
29.
?5.
28.
19.
18.
IP.
19.
19.
18.
78
98
17
51
16
75
90
85
39
63
00
05
56
72
XP
76
18
3'
58
67
59
71
01
12
21
80
65
19
48
9.22
7.19
9.43
9.43
7.31
7.^5
8.58
8.59
3.64
5.04
5.03
3.64
4.46
4.89
4C.SVSP
20.8*
19.27
»2.32
23.47
21.93
16.39
19.96
18.52
13.55
18.02
15.56
9.95
13.85
13.48
N02 *N02 TADJ HC/Nt!X NO»R
.15 10
.18 11
.15 10
.15 10
.17 11
.15 10
.16 10
.16 10
.17 11
.16 10
.17 11
.13 9
.18 11
.18 11
P3MAXC
0.92
0.92
0.83
0.85
0.8?
1.19
1.13
1.17
0.71
0.70
0.75
0.96
0.86
0.89
HfWX
.67
.75
.58
.89
.86
.83
.78
.84
.85
.81
.79
.57
.80
.89
.1
.8
.6
.5
.4
.3
.7
.8
.P
.7
.2
.0
.6
.8
03TMAX
113.
109.
87.
92.
105.
1'5.
106.
117.
160.
132.
136.
180.
148.
138.
o«NMM<
.29
.23
.26
.29
.25
.24
.20
.29
.22
.19
.18
.31
.24
.29
0. 2.
0. 2.
0. •>.
0. 2.
0. 1.
0. 2.
0. 1.
0. 2.
3. 2.
0. 2.
2. 2.
0. 2.
3. 1.
3. 1.
03nnsF
168.9
170.9
163.6
151.1
164.3
204.3
197.1
213.9
120. T
134.0
138.4
141. ">
148.9
149.3
NOR'TF
19.88
18.66
20.58
22.12
20.62
15.53
18.18
15.97
12.33
17.21
14.77
°.84
12.85
11.4?
16
00
18
17
92
07
99
07
16
05
13
00
95
96
HfPf.
.08
.11
.09
.14
.16
.n
.09
.11
.13
.17
.17
.29
.11
.27
30.5
29.5
'8.6
27.5
29.7
32.0
30.1
30.5
43.2
35.2
36.4
42.1
?7.4
33.9
36.10
29.85
57.14
52.08
52.36
19.46
?0.5P
26.25
19.84
40.00
27.25
12.18
22.42
20.66
MCT75
45.
47.
41.
34.
42.
66.
46.
54.
64.
44.
59.
104.
59.
63.
= l»r»p
36.9
32.1
37.8
*9. «?
37.1
46.2
43.4
47.6
27.5
29.5
29.3
3».4
31.9
31.6
STSINLESS STE*L
RUN NO
55
56
51
52
58
62
61
66
57
53
54
63
60
67
PHHH
71.
67.
61.
62.
64.
79.
75.
69.
107.
88.
90.
114.
100.
97.
XTIHF
20.
24.
13.
13.
15.
33.
23.
27.
36.
19.
25.
50.
33.
35.
N02XT
'.8.
27.
22.
21.
?4.
36.
31.
33.
47.
31.
38.
52.
43.
40.
03XT
92.
95.
7".
80.
91.
93.
83.
90.
124.
113.
111.
130.
115.
103.
A-21
LOCKHEED MISSILES & SPACE COMPANY, INC.
-------
LMSC-D406484
Covariate Adjusted Data
OR
?'JN NO
•x
4
4.5
10
11
1 ?
12.5
6
5
9
7
3
o»
'UN NO
3
4
4.5
10
H
12
12.5
6
5
9
7
8
I r.iNiL
s/v s
2
2
2
1
1
1
1
?
7
I
1
1
ir.iNU
V 0 7 T M
78.
75.
76.
78.
87.
88.
86.
130.
99.
m.
112.
115.
ORIGINAL
?UN NO
3
It
4.5
10
11
12
12.5
6
5
9
7
ft
HCT50
99.
99.
102.
108.
108.
114.
114.
134.
128.
168.
148.
150.
ORIGINAL
RJN NO
3
4
4.5
10
11
12
12.5
6
5
9
7
8
PtfHM
73.
65.
58.
72.
66.
73.
71.
108.
90.
115.
100.
97.
ALUMINUM
P^C Cl«=AN
2 2
2 1
2 1
2 2
7 7
2 1
2 1
I 2
I 1
1 2
1 1
1 1
ALUMINUM
NT2D05 n
141.
137.
121.
139.
143.
166.
164.
173.
128.
1.69.
157.
139.
ALUMINJM
HCT75 H
125.
130.
130.
140.
138.
137.
141.
171.
181.
198.
194.
194.
ALUMINUM
XT IMC
39.
42.
" 45.
42.
45.
47.
46.
54.
56.
70.
55.
62.
HCINTT NHX
3.17
3.05
7.90
7.96
3.03
7.9*
2.83
7.81
7.98
?.°4
7.84
2.9?
3MAX°
21.60
79.59
78.8?
27.93
35.46
34.1?
28.01
18. 93
18.08
21.37
27.94
26.74
CMAXR
75.06
24.15
23.64
74.75
76.53
27.70
28.17
19.31
1 8. 0?
23.64
72.22
77.77
ND2XT
37.
37.
37.
36.
37.
41.
41.
56.
44.
61.
58.
53.
1.52
1.54
1.41
1.51
1 . 43
1.51
1.67
l.«59
1.33
1.49
1.4?
l."»5
n3AVf",°
4.96
5.89
5.25
5. *>*5
5.7?
5.79
4.95
3.35
2.86
2.89
3.41
3.51
HCAVG»
15.34
15.60
15.75
13.03
13.76
13.18
12. ?0
11.96
11.38
8.42
9.30
9.35
03 XT
87.
77.
71.
95.
82.
86.
87.
121.
101.
121.
110.
105.
N02 *N02 Tflru HT/NOX M07R
.22 14.5 9. 2.09 15.8?
.29 19.8 17. 1. 08 15.50
.29 20.6 21. '.05 13.40
.31 20.5 20. 1.9f> 16.64
.2? 15.4 H. 2.1? 15. 3P
.11 7.3 -p. 1.97 14.58
.36 19. •» 18. 1.69 15.06
.31 n.5 78. 7.04 11.31
.12 9.0 0. 2.74 9.07
.33 22.1 35. 1.97 7.P7
.7? 15.5 16. 7.00 10.2?
.19 14.0 12. 7.1" 8.57
33MftXC T3TM4X H3DOS? H" = f. -ICT75
0.84 128. 138.? .0? 55.
1.0? 120. 163.3 .38 65.
I. 00 117. 154.6 .10 61.
1.11 138. 199.6 .18 7«.
1.11 1?.7. 192.9 .14 78.
1.16 133. 19?. 5 .14 97.
0.98 137. 169.4 .15 89.
0.86 175. 136. * .16 99.
0.7B 157. 120.3 .35 89.
0.97 191. 171.3 .30 134.
0.87 164. 128.4 .29 119.
0.94 167. 144.2 .29 122.
AlOMAX PANMAX NORATE NO'OF P'OP
.66 .57 14.97 31.1 31.?
.57 1.04 12.97 31.7 37.6
.47 1.00 12.08 30.4 39.7
.57 .04 11.99 *2.2 45.7
.61 .15 11.36 34.1 45.6
.49 .35 11.51 35.6 41.3
.48 .24 11.95 33.8 34.6
.70 .13 10.48 37.4 37.1
.72 .35 9.81 31.2 2^.2
.47 .07 7.34 40.0 29.5
.46 .11 9.18 37.4 31.4
.55 .11 9.2? 34.? 36.0
A-22
LOCKHEED MISSILES & SPACE COMPANY. INC.
-------
RELATIVE HUM 10ITY EFFECT
ALUMINUM
RUN S/V SPEC P.P. CR
~7T LOW FULL 55
76 LOW FULL 22
50
100 150 200
TIME (MIN)
250 300
RELATIVE HUMID ITY EFFECT
ALUMINUM
RUN S/V. SPEC P.P. CF)
69485 HIGH FULL 5$
89 HIGH FULL 19
. 0,
100 150 200
TIME (MIN)
250 300
RELATIVE HUMIDITY EFFECT
ALUMINUM
. RUN SJV SPEC D.P. CFI
828,84 HIGH CUT 55
88 HIGH CUT 14
RELATIVE HUMIDITY EFFECTS
STAINLESS STEEL
RUN SN SPK P.P. CF)
51, 52. 58 HIGH FULL 55
59 HIGH FULL 12
50
100 150 200
TIME (MIN)
250 300
100 150
TIME (MIN)
A-23
LOCKHEED MISSILES & SPACE COMPANY. INC.
-------
IMSC -
RELATIVE HUMIDITY EFFECT
STAINLESS STEEL
RUN S/V SPEC D. P. CF)
61 & 66 LOW - FULL . 55
64 LOW FULL 22 —
65 LOW FULL -20
RELATIVE HUMIDITY EFFECT
STAINLESS STEEL
RUN s/v SPEC p.p. rn
60~O7 LOW "CUT K
68 LOW CUT -12
50
100 150 200
TIME (MINI
250 300
100 150 200 250
TIME (MIN)
300
RELATIVE HUMIDITY EFFECTS
TEFLON
RUN S/V SPEC P.P. CF)
46 LOW FULL 55
47 LOW FULL 17^-
49 LOW FULL . -20
RELATIVE HUMIDITY EFFECT
TEFLON
RUN S/V SPEC D.P. CF)
41 LOW CUT 55
48 LOW CUT 15
50
100 150 200
TIME (MIN)
250 300
100 150 200
TIME (MIN)
250 300
LOCKHEED MISSILES & SPACE COMPANY. INC.
-------
LMSC-D406484
BUTANE/NOX
ALUMINUM
Q2.0
I
Z
o
u
BUTANE
NO
NO,,
OZONE
— 3.0
2 2.0
UJ
o
U
^3.0
12.0
ut
o
u
50 100 150 200 250 300 350
TIME (MIN)
BUTANE/NOy
ALUMINUM
RUN S/V SPEC
79B LOW FULL
3.0
2 2.0
1.0
o
u
0 50 100 150 200 250 300 350
TIME (MIN)
BUTANE/NO.
ALUMINUM
RUN S/V
87B HIGH
SPEC
CUT
0 50 100 150 200 250 300 350
TIME (MIN)
BUTANE/NOy
ALUMINUM
RUN S/V SPEC
80B LOW CUT
0 50 100 150 200 250 300 350
TIME (MIN)
A-25
LOCKHEED MISSILES & SPACE COMPANY. INC.
-------
LMSC-D406484
BUTANE/NOy
TEFLON
RUN S/V SPEC
95B HIGH FULL
Q.
3.0
^2.0
BUTANE
NO
NO-
u
O
U
1.0
-3.0
g2.0
O
u
?3.0
z
P. 2.0
UJ
O
u
50 100 150 200 250 300 350
TIME (MINI)
BUTANE/NOX
TEFLON
RUN SA SPEC
94B LOW FULL
i
3.0
Z
? 2.0
UJ
O
u
BUTANE/NO
TEFLON
Y
RUN
96B
HIGH
SPEC
CUT
0 50 100 150 200 250 300 350
TIME (MIN)
BUTANE/NOX
TEFLON
RUN VV SPEC
93B LOW CUT
50 100 150 200 250 300 350
TIME (MIN)
50 100 150 200 250 300 350
TIME (MIN)
A-26
LOCKHEED MISSILES & SPACE COMPANY, INC.
-------
LMSC-D406484
Appendix B
STATISTICAL ANALYSIS
Prepared by Ken W. Last and Michael W. Reeder
1. 0 INTRODUCTION AND OVERVIEW
1. 1 The analysis has been complicated by a number of problems commented
upon separately below. There have been a number of opportunities to choose
among several alternative methods of analysis. In the body of the report we
have tried to show not only what was done (in the way of analysis), but have
also tried to indicate the reasoning that led to the choices selected.
1. 2 The analysis establishes that initial conditions are important. In par-
ticular, initial HC and % NO2 were found to be most effective in standardizing
the data. NO., also contributed when the data was viewed in a Multivariate-
A.
all the parameter taken together fashion, or when certain materials were con-
sidered alone (i.e. , Pyrex). One might well speculate that different correc-
tion factors for initial conditions should be developed for cut spectrum vs.
full spectrum or for the various materials, however, there simply is not
sufficient data to pursue this point. The corrections for initial conditions
which were developed are given in Table B3-1.
1.3 Among the effects, that of spectrum dominates virtually every para-
meter. S/V is also important in a number of parameters, while cleaning is
less important. These effects are not consistent across materials although
the spectral effect is so large that it is clearly visible even in the midst of
the interactions with materials. The analysis across materials is summarized
B-l
LOCKHEED MISSILES & SPACE COMPANY
-------
LMSC-D406484
in Tables B4-5 thru B4-9. Note that cleaning has a substantially greater effect in stain-
less steel than in any of the other materials. One other notable feature of
the data is the repeated occurrence of a significant Material-Spectrum -S/V
interaction: interpreted as meaning that the effect of spectrum depends upon
the S/V ratio, however, the form of this dependence not consistent across
materials. This results generally from an increased sensitivity under cut
spectrum conditions. When the runs are full spectrum, the system is "fast"
and subsidiary effects are washed out, whereas, when the system is run with
cut spectrum the system slows down and other effects have an opportunity
to exert their influence.
After applying several tests which would indicate drift, the conclusion drawn
is that there is no compelling evidence of drift either within each block of
material (excluding the original set of aluminum runs) or from material
block to material block.
1. 4 There are very strong interrelationships among the 23 parameters
which tie together NO , ozone and HC. Thus, when the system is "fast" it
Jt
is fast with respect to all three. Substantially fewer than 23 parameters
are adequate to describe the system. The 21 parameters involving NOV
.X
ozone and HC can be reduced to four which contain the great majority of
the information represented by the original 21.
Of the four "principal components" two stand out in particular. The first of
these represents "time" — i. e. , time to max value, time to decay a fixed
fraction, etc. In addition, "time" also represents rates as well. The second
component represents "dose", the integral under the curve. These two
B-2
LOCKHEED MISSILES & SPACE COMPANY
-------
LMSC-D406484
principal components are quite highly correlated which indicates that con-
centration patterns are consistent with the time effects in response to changes
in chamber conditions. Two other components complete the set, but these
are both less consistent and more obscure in their interpretation. The analy-
sis reported in Table B5-15. provides a summary of the effects of S/V, spec-
trum and cleaning material by material which is, in effect, an overview of
the results for the 21 parameters (excluding ALD MAX and PAN MAX).
1. 5 There are a number of anomolous characteristics of the data; runs which
seem inconsistent. In particular, run 25 on pyrex and run 4-1 on Teflon stand-
out as unusual. Also run 42 on stands out with respect to PAN MAX
These runs can have a profound influence upon the analysis and we have
tended to analyze without deleting them and then to conduct some informal
analyses which indicate what their effects may be.
1. 6 We have been particularly troubled by the possibility that the results
of the program may be confounded by the presence of drifts introduced by
aging, instrumental errors, etc. The form of the experimental design is
such that the program is especially vulnerable to such drifts. In particular,
a steady trend would tend to be confused with a S/V effect.
2.0 BACKGROUND
2. 1 The Experimental Design
Three variables-S/V ratio, spectra, and cleaning method-each at two levels,
were combined into a complete 2x2x2 factorial experiment for each of four
materials (aluminum, pyrex, teflon, stainless steel). In addition, upon com-
pletion of these materials, a complete rerun of the aluminum runs was con-
ducted.
B-3
LOCKHEED MISSILES & SPACE COMPANY
-------
LMSC-D406484
Certain of the combinations were repeated to provide a measure of repro-
ducibility, while others were repeated at the convenience of the experimenter.
The order in which the experiments was conducted was not randomized. The
possible consequences of not running the experiments in random order are
discussed in Section 2. 7 below.
2. 2 The Data
The basic data for each run consists of a set of initial conditions (initial HC,
NOX, NO2) together with (nearly continuous) graphs of NO , NO?, Ozone,
Propylene, Acetaldehyde, and PAN. From these graphs are derived a set
of 23 parameters as defined in Section 3 of the body of the report. These
parameters are intended to summarize the time-wise behavior of the photo-
chemical system. Indeed, through the use of the parameter set it is pos-
sible to generate graphs which appear to be good approximations to the origi-
nals. Note, however, that within the framework of the present statistical
analysis no systematic effort has been addressed to evaluating better sum-
mary measures or toward modeling the chemical system in a direct fashion.
Rather, the major emphasis has been directed towards analysis of the para-
meters as defined. Note that, of the 23 parameters, 13 are strongly involved
with time.
2. 3 Covariates
While a careful effort was made to provide each experimental trial with the
same starting conditions, so that observed changes could be attributed only
to the deliberately introduced variations in the experimental design, some
variation in these conditions did occur.
B-4
LOCKHEED MISSILES & SPACE COMPANY
-------
LMSC-D406484
In order to remove this extraneous source of variability, an analysis of co-
variance was performed. The intent of the analysis is to adjust each'trial
to standard initial conditions (using a linear relation) in a manner which best
aides in clarifying the other effects.
Early in the program it was observed that initial NO^ could be used as a
basis for standardizing "time" within the system — that is, sliding the con-
tinuous graphs so as to bring the early NC^ values into coincidence, seems
to bring replicates into coincidence across the entire time axis as regards
the occurrence of peaks, etc. Such an adjustment to the same percentage
content of NO, in the NO has been previously suggested after both theoretical
and experimental consideration of the photochemistry (Dimitriades, Ref. 14
and Niki, Ref. 7). The results incorporating a time adjustment were reported
thereafter as the basic parameter set.
In this analysis, we have returned to the unadjusted data and let the co-
variance analysis "find" the adjustment. This analysis resulted in compara-
ble data as is shown in Table B2-1. This table compares the standard de-
viations calculated from replicates for each data set. Generally, it can be
seen that the covariate set is superior. One must recognize that the co-
variance corrections are limited by the very nature of the derived parameters.
Some of the parameters represent pure times and provide no particular prob-
lem. Others, however, involved time nonlinearly. For example, the calcu-
lation of dose involves integrating under a curve until a fixed time index and
the consequence of sliding the index can be expected to influence the derived
parameter in a nonlinear fashion.
B-5
LOCKHEED MISSILES & SPACE COMPANY
-------
TABLE B2-1
IMSC-D406484
PARAMETER
Est. Std. Dev.
As Initially
Corrected
1*
2
3
4*
5*
6
7
8
9
10
11
12
13*
14*
15
16
17*
18
19
20
21
22
23
N02 Rate
N02 Time Max
NDp Dose
Ozone Max Rate
Ozone Avg. Rate
Ozone Max. Cone.
Ozone Time Max.
Ozone Dose
HC Final Cone.
HC Time .75
HC Time .5
HC Time .25
HC Max Rate
HC Avg. Rate
Max Aid.
Max PAN
NO Rate
N02DF
Ozone DF
N02PWAM
Crossover Time
2-21
7-21
.0091
7-35
8.43
.0057
.0115
.033
8.05
8.51
.039
8.34
6.85
6.76
.0057
.0062
• 135
•039
.0104
1.93
1.68
4.83
4.18
2.91
4.87
Est. Std. Dev.
Using Covariate
Correction
.0082
4.54
8.02
.0108
.0145
.038
7.6l
7.65
•039
7.01
6.13
6.30
.0052
•0053
.0817
.042
.0070
1.90
2.10
4.72
4.19
2.50
4.63
Based on reciprocals,
B-6
LOCKHEED MISSILES & SPACE COMPANY
-------
LMSC-D406484
The point is, ideally, the covariates would be used to adjust the graphs to
standard conditions and then the derived parameters would be read from the
graphs. Actually the procedure used attempts, linearly, to adjust the para-
meters and cannot help but be less effective than it might be otherwise.
One of the judgemental decisions which had to be made in performing this
analysis was whether or not to conduct the multivariate analysis of variance
and covariance on the data as originally adjusted or upon the raw data (with
the attendant risk of nonlinearities). The analysis reported in
Section 3, led to the expectation that the covariance on unadjusted data would
be fairly successful, and this together with the favorable comparison in
Table B2-1, has led us to base the bulk of the present analysis upon the raw
(unadjusted) data, and to use covariance analysis to introduce any corrections
needed.
2. 4 The Unbalanced Design and Its Consequences
One major advantage in analyzing data from a "balanced" experimental de-
sign is the orthogonality of the resultant estimates: each estimate stands
alone and is not changed, when the underlying model is changed. For example,
the estimate of the effect of changing the S/V ratio is the same whether or not
the model includes an estimate of the effect of changing the method of clean-
ing. With unbalanced data this feature is lost and generally the estimates
depend upon what terms are in the model. Moreover, the sums of squares
attributed to particular effects (used in forming tests of significance), de-
pend on the order in which they are considered. As a consequence, the test-
ing of hypotheses concerning the significance of the various effects is greatly
complicated.
B-7
LOCKHEED MISSILES & SPACE COMPANY
-------
LMSC-D406484
An "orthogonalized" analysis has been reported in the body of the report.
This analysis avoids the complexities of an unbalanced analysis by averag-
ing replicates and then proceeding as if only a single result were available
at each trial. Such a procedure sacrifices some efficiency and invalidates
(to a moderate extent) the formal tests of hypothesis. It does, however,
enjoy the important advantage of simplicity and ease of interpretation. So
much so, that in this Appendix results are again presented in this form, in
addition to the results from the more formally correct unbalanced analysis.
2. 5 Principal Comments
One of the basic goals of the current statistical analysis has been to clarify
and illuminate the data. It is difficult (if not impossible) to visualize 23
dimensions. Yet, examining the variables one at a time can be misleading
since it fails to exploit the rather strong system structure from which the
parameters are all derived.
A fairly common technique used to reduce the dimensions without losing the
structure is the technique of principal components. Here the object is to
"economize in the number of variates" by seeking linear combinations of the
original variates which account for most of the variation.
Section 5 presents the results of such a reduction in dimensions, from the
original 23 down to four.
2. 6 Outliers
This data set is rather sensitive to anomalies since the bulk of the trials
were performed only once.
B-8
LOCKHEED MISSILES & SPACE COMPANY
-------
LMSC-D406484
There are a number of trials where the data as represented by one or more
parameters might be considered unusual. For example, the unusually high
values of PAN max on Run 42 (teflon).
Our policy throughout the program has been the following: If the experimenter
during the course of a run felt a physical or chemical grounds that the results
were invalid, he was free to supress the data. Once, however, the data was
reported for analysis, then that data could be modified only upon the basis of
formal justification, i. e. , it was presumed valid unless proved otherwise.
As a consequence of this policy, we have been reluctant to throw out data
even when the statistical analysis makes us suspicious that it is, in some
sense, anomalous.
Note, however, that in reporting the results of the analysis of the four derived
principal parameters in Section 5. 3, two runs (No. 25 on Pyrex and No. 41
on Teflon) have been suppressed. In each case these were extremely
slow runs with cut spectra. Sequentially in time, Run 25 was the first cut
spectrum run for Pyrex while Run 41 was fifth (final) run for cut spectrum
on Teflon.
In both materials the cut spectrum runs appear to exhibit a time trend —
becoming faster in Pyrex; becoming slower in Telfon.
We can only speculate about the source of the evident discrepancy of these
runs. Note, however, that no explanation has been advanced as to why any
"drift" such as lamp or filter aging, or instrumental bias should cause a
striking increase in one material and an opposite effect in the other —
B-9
LOCKHEED MISSILES & SPACE COMPANY
-------
LMSC-D406484
particularly when nothing of a similar character is seen in the other materials.
Moreover, as is seen in Section 5. 3, the data, after suppressing these runs,
exhibits substantial regularity as regards the effects of S/V, spectrum, and
cleaning. One possibility is that the underlying photochemical reactions be-
gin to undergo a basic change in character when things get slow enough and
that other factors which are not even being considered in the variable set be-
come important.
2. 7 Time Trends
In forming the experimental design for this program it was found expedient
to omit the usual randomization of the order in which the experiments were
to be performed. This was a consequence of the fact that changing materials
involves rebuilding the smog chamber and that changing the surface to volume
ration (for a given material) involves a partial rebuilding and resealing of the
chamber. Initially this led to a design in which all runs for a given material
were to be performed in a block. Moreover all runs for a given surface to
volume ratio would also be run as a block. Subject to these constraints, the
following randomization steps could still be performed.
1. Randomize the order in which the materials were run.
2. Within each material block, randomize the order in which the
S/Vs were run (i. e. , all high runs then all low or visa versa).
3. Within each material-S/V block, randomize the order in which
the four runs comprising the block were run.
In addition the very limited number of repeats were scheduled so as to repre-
sent each of the following conditions:
B-10
LOCKHEED MISSILES & SPACE COMPANY
-------
LMSC-D406484
a. An immediate rerun.
b. A rerun without the chamber seal being broken, i. e. , a rerun
within the same S/V block, but not an immediate rerun.
c. A rerun within the same material block, but after a S/V change,
and finally,
d. A rerun after a material change.
During the initial runs (which were, coincidently, high S/V full spectrum
runs) it was noted that it seemed to take several runs to achieve stable be-
havior and that this problem might be attributed to a "virgin surface effect".
This being the case it seemed prudent to start the runs with all the material
in the chamber exposed to the highest light intensity. Thus, the random-
ization was further restricted in that within each material block the first
S/V block was invariably at high S/V and, within each high S/V block the
initial runs were at full spectrum.
The end result has been that the experimental design is quite vulnerable to
drift or other time trends. If such trends were present they would tend to
be confused (confounded) with the estimated S/V effects within materials and
with the material to material comparisons. This weakness in the design was
recognized and a decision was made to proceed using insofar as possible,
experimental controls (calibrations, frequent Kj checks, etc.) to prevent
any drift from arising.
Fortunately, in the course of the program, the experimental procedures
achieved an increase in efficiency which permitted an increased amount of
replication to the extent that the replicates can be used to form a judgment
as to the existance of drift during the program.
B-ll
LOCKHEED MISSILES & SPACE COMPANY
-------
LMSC-D406484
There are a number of ways in which an analysis might be undertaken which
addresses the question of drift.
One possibility is to undertake, material by material, a stepwise regression
to examine which variables are important in the sense that they provide a
good fit to the observed data. This was done and revealed that run sequence
provided, in a number of cases, a. superior fit to that which resulted when
this variable was excluded. In addition, it was noted that whenever run
sequence entered the model, S/V was omitted. Also, it was found that when
the models were revised to force S/V into the model and to add run sequence
only if it significantly enhanced the fit beyond what S/V had contributed, that
no significant additional enhancement resulted.
Thus, this analysis was inconclusive in that we were left in a position of
finding either run sequence or S/V (but not both) to be important. Rather
than pursue this analysis further, an alternate has been used.
The hypothesis to be tested concerns the presence of a steady drift over run
sequence. If such drift were present, then runs which are duplicates (except
for sequence) might be expected to differ by an amount which grows steadily
as a function of the separation overtime. Our attention is then focused upon
an examination of the differences between replicate pairs as a function of
run sequence separation.
Note that the time separations of replicates tend to fall into three basic
groups:
B-12
LOCKHEED MISSILES & SPACE COMPANY
-------
LMSC-D406484
(1) Immediate reruns - (A = 1). There are nine such
4-4A Aluminum
7-8 Aluminum
10-11 Aluminum
12 -12 A Aluminum
34-35 Teflon
43-44 TeHon
51-52 Stainless
53-54 Stainless
55-56 Stainless
(2) Sequence numbers differing by more than 1 but less than 20 (1<4 < 20).
There are ten such*.
27-33 Pyrex
26-39 Teflon
51-58 Stainless
52-58 Stainless
60-67 Stainless
61-66 Stainless
70-82 Aluminum (Rerun)
70-84 Aluminum (Rerun)
69-85 Aluminum (Rerun)
82-84 Aluminum (Rerun)
*A"complication in application of these tests has been the triplets formed by
runs 51, 52 and 58 and by runs 70, 82 and 84. Clearly we are not justified in
using three pairs (one of the three differences is simply the sum of the other
two differences). We have avoided this problem by dropping the middle value,
thus leaving in the replicate, the difference involving the largest run sequence
separation.
B-13
LOCKHEED MISSILES & SPACE COMPANY. INC.
-------
LMSC-D406484
(3) Long term replicates: A complete rerun of the aluminum block.
There are two basic data sets. First the data incorporating the NO2 time
adjustment, and second, the unadjusted data after incorporation of the co-
variate corrections. In addition, as will be seen below, we have some reason
to be less confident of the initial runs on aluminum and, the analysis has been
redone, dropping these pairs.
In the analysis, we concentrate on the differences between duplicate pairs
since this provides a direct basis for comparison. In the absence of drift,
these pairs should be distributed around a central valiE of zero. If there is
drift then the pairs separated further apart in time (or sequence) could be
expected to center about a value, different than zero and increasing in
absolute magnitude as the separation increases. Note, however, that the
long term replicates group 3 all involve only one material, aluminum, and
that each set of runs, (initial aluminum, replicate aluminum) were fairly
compact in time. One might anticipate in this case a different kind of drift
than which might occur during a single series. Thus, we distinguish two
types of drift, drifts within a compact series on runs on a given material,
and drifts from material to material, or series to series.
The drifts within materials can be studied using the shorter term replicates,
groups 1 and 2, while the drifts across materials or series can be studied by
comparing the mean values across materials and especially between the
initial aluminum set and the final replicate aluminum set.
B-14
LOCKHEED MISSILES & SPACE COMPANY, INC.
-------
LMSC-D406484
2. 7. 1 Drifts within Material
If there is a drift within material then the replicates with longer separations,
group 2, should be centered around a different value than those of group 1.
Hence we test the hypothesis that both sets are centered about the same
central value. The test used is the rank sum test using the value of T
defined as follows: Arrange the two samples (of differences) in order of size,
and assign rank scores to the individual observations, score 1 for the
smallest, 2 for the next smallest, etc. Then T is the sum of the tanks of
the observations in the smaller of the two sets. Using tables, (Reference B-l),
we reject the hull hypothesis (no drift) if the calculated score is significantly
large or significantly small where the critical values are selected to provide
a probability of approximately 0. 05 of falling in each tail (10% overall).
The rank sums calculated in this manner are shown in Tables B2-2, B2-3
and B2-4.
For the adjusted data only one parameter (13-HC Max Rate) qualifies as
significant. For the covariate corrected data three parameters qualify as
significant. These are: 1-NO., Rate; 21-Crossover Time; and 23-Ozone
Time Max to Crossover Time. For the covariate corrected data, omitting
early aluminum runs, two parameters (19-Ozone DF and 21-Crossover Time)
qualify as significant.
Here, after applying a test which would declare significance about 10% of
the time when no time trend is present, we have found in the case of the
covariate corrected data, three parameters which may be displaying drift
or trend. This is just about what would be expected to occur by chance if no
trend were present in any parameter.
B-15
LOCKHEED MISSILES & SPACE COMPANY. INC.
-------
IMSC-D406484
TABLE B2-2
Rank Sum and Sign Test Results for Adjusted Data
PARAMETER
Rank Sum
+
Sign Test 0
—
PARAMETER
Rank Sum
+
Sign Test 0
—
PARAMETER
Rank Sum
+
Sign Test 0
-
PARAMETER
Rank Sum
- +
Sign Test 0
-
1
47
11
4
7
^9
5
10
13
39-5
. 4
11
19
52
5
1
9
2
49.5
3
1
11
8
51
4
1
10
Ik
61.5
12
3*
20
57
7
8
3
58
7
8
9
52.5
3
22
10
15
62
6
9
21
63.5
4
1
10
if
66
9
6
10
48.5
4
1
10
16
56
7
1
7
22
58.5
5
3
7
5
. 59
11
4
11
50.5
2*
1
12
17
57
9
6
23
45
5
3
7
6
58
7
1
7
12
53
2*
1
12
18
49
8
7
Critical values for rank sum are 4l and 71
Critical values for sign test are 2 when N = 12, 13 or l4
Critical values for sign test are 3 when N = 15
B-16
LOCKHEED MISSILES & SPACE COMPANY
-------
TABIE B2-3
Sign Test Results for Covariate Corrected Data
PARAMETER
Rank
+
Sign Test 0
*
PARAMETER
Rank
+
Sign Test 0
-
PARAMETER
Rank
+
Sign Test 0
—
PARAMETER
Rank
+
Sign Test 0
-
1
72*
k
0
11
7
52.5
6
l
8
13
59
8
1
6
19
^5
9 '
0
6
2
55-5
3
2
10
8
5^-5
10
0
5
ih
c|i
5
0
10
20
58.5
8
0
7
3
*5-5
7
2
6
9
53
3
2
10
. 15
55
6
0
9
21*
73*
3*
0
12
k
56
6
1
8
10
l*.5
3
1
11
16
57-5
8
2
5
22
55
5
U
6
5
57
4
2
9
11
5^.5
3
22
10
17
E^h
|i
1
10
23*
40*
6
3
6
6
59
7
fy.
ij.
12
58
3
2
10
18
^9-5
7
2
6
Critical Values for Rank Sum are **•! and 71
Critical Values for Sign Test are 2 when N - 12, 13, ill-
Critical Values for Sign Test are 3 when N = 15
B-I7
LOCKHEED MISSILES & SPACE COMPANY
-------
TABLE B2-U
Covariate Corrected Data (Original Alum Omitted)
1
15-5
3
0
8
2
22
2
2
7
3
27
4
2
5
k
23
4
1
6
5
25
2
2
7
6
21
5
3
3
PARAMETER
Rank Sum
Sign Test 0
PARAMETER 7 8 9 10 11 12
Rank Sum 21.5 20.5 23.5 28.5 23.5 17
+ 37 2 23 2
Sign Test 0 1 0 100 0
7 4 8 98 9
PARAMETER 13 l4 15 l6 17 l8
Rank Sum 2k 22.5 23.5 22.5 25 26.5
4
Sign Test 0000112
5
PARAMETER
Rank Sum
Sign Test 0
Critical values for rank sum are 15 and 33-
Critical values for sign test is 1 for N = 9, 10 or 11.
13
2k
6
0
5
19
35*
7
0
4
14
22.5
3
0
8
20
21
5
0
6
15
23-5
4
0
7
21
14*
2
0
9
16
22.5
6
1
k
22
22.5
4
2
5
17
25
3
1
7
23
31
3
3
5
BJ8
LOCKHEED MISSILES & SPACE COMPANY
-------
LMSC-D406484
Next, we test the null hypothesis that the central value of the differences is
zero. If this were true, then there would be equal chance for each difference
to be either positive or negative (a difference of zero is excluded). Letting
r represent the smaller of the counts of positive or negative differences
then the null hypothesis is rejected if r is too small. In particular using
Table A-lOa (Ref. B-l) we select r such that there is approximately a 10%
chance of falling in the critical region if the null hypothesis is true.
The counts are also shown in Tables B2-2 to B2-4. For the adjusted data
three parameters show significant counts. These are 11-HCT. 5, 12 HCT. 25,
and 14-HC Avg. Rate. For the covariate corrected data, one parameter
(21-Crossover Time) shows a significant count. For the covariance corrected
data, omitting early aluminum, no parameter shows a significant count. As
was the case with the rank sum test, the number of parameters qualifying
as significant is approximately what would be expected to occur by chance if
no trend were present.
In only one case (21-Crossover Time) do both tests qualify. In this case the
majority of the differences are negative implying a down trend with sequence.
However, the rank score is significant on the high end implying that the
differences show an increasing trend, so that the two results are inconsistent.
All in all, the conclusion drawn is that there is no compelling evidence of
drift. Note also that the Pyrex and Teflon runs at cut spectrum seem to show
a time trend (See Fig. B5-2 and B5-3 for both "time" and "dose" parameters).
However for Pyrex the trend appears to be for faster runs with increasing
sequence number while for Teflon the trend appears to be just the reverse
(towards slower runs with increasing sequence number).
B-19
LOCKHEED MISSILES & SPACE COMPANY. INC.
-------
LMSC-D406484
2. 7. 2 Material to Material Time Drift Analysis
Next, we consider whether there is any indication of a long term drift from
material block to material block. First note that there is .no evident pattern
in the block means arranged in sequential order. Table B2-5 presents the
means parameter by parameter based on the orthogonalized analysis. In
that table there are 17 parameters which appear to follow a steady upwards
or downward progression starting with the original aluminum block and
continuing thru Pyrex (the second material run) and Teflon (the third). This
pattern is not continued thru the stainless steel runs, where, of the 17,only
two parameters continue the trend. These are 4-Ozone Max. Rate and 14-Max.
Aid.
Of greater interest is the comparison between the original aluminum block
and the rerun block. This is material contrast two reported in Table B4-4.
From that table it can be seen that for 8 of the parameters a significant
difference is found and in 6 additional parameters a significant Material-S/V
interaction is found. Evidently these two sets are not consistent with regard
to their individual means or the S/V effect or (In two cases) both. This can
be interpreted as evidence of drift. However, a comparison of the two sets
indicates substantial inconsistencies between points which would be replicates
except for drift.
Now that we have seen how highly correlated the various "time" parameters
are, it can be seen that to a large extent these tests are redundant -- testing
for the same thing over and over --so that it is no surprise that a small
number of the tests show significance.
B-20
LOCKHEED MISSILES & SPACE COMPANY. INC.
-------
LMSC-D406484
Table B2-5
MEANS BY MATERIAL
TIME -~
Parameter
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
Alum
(Original)
11.65
104.2
151.5
25.4
4.15
0.95
157
147
0.19
97.6
131.2
166.2
23.1
11.66
0.597
0.37
10.81
34.1
32.9
87.2
57.7
46.2
99.4
Pyrex
10.00
130.9
178.0
24.7
3.60
0.93
189
126
0.26
116.6
156.5
188.8
23.6
10.13
0.696
0.14
8.55
39.2
27.8
102.8
78.2
57.5
114.5
Teflon
8.63
154.2
219.0
21.3
3.12
1.03
230
131
0.26
124.4
179.5
218.2
15.9
8.18
0.752
0.29
7.20
48.7
28.6
124.9
81.6
69.9
147.5
S.S.
25.38
70.20
152.8
21.2
5.68
0.90
137
157
0.15
64.2
96.8
133.4
23.9
16.16
0.758
0.23
14.36
34.5
34.1
85.8
38.2
35.8
102.6
Alum
(Repeat)
13.12
105.8
176.1
24.7
4.08
1.00
172
156
0.16
93.6
132.9
165.9
21.8
11.52
0.870
0.35
8.71
39.0
34.3
99.4
56.2
49.6
116.2
B-21
LOCKHEED MISSILES & SPACE COMPANY. INC.
-------
LMSC-D406484
As regards long term drift -- based on looking for trends from material
to material and especially as regards the original versus the rerun aluminum,
there are some important conclusions: 1. There is a definite change in
parameter 15 (Max Aid) due to a change in the analytical method. 2. There
is a similar change in parameter 16 (Max PAN) for similar reasons. In
particular the values reported for runs 3, 4 and 4A are very different for
corresponding runs on the aluminum reruns. Moreover, the two sets original
versus rerun aluminum are generally inconsistent. 3. There are inconsistencies
(already reported in the Phase II final report) between the original and rerun
of aluminum, which are definitely not resolved by the introduction of the
covariate corrections. This latter point is further examined in the next
paragraph.
A tentative explanation is that during the initial aluminum runs there was a
run sequence effect, equivalent to an experience factor. It is our considered
judgment that the early aluminum runs should be dropped from consideration
on the basis that a learning process was taking place and that several changes
in procedure were instituted during these runs. It is felt (but cannot be
established statistically) that the system had stabilized by the time these runs
were completed.
It must,however,be acknowledged that the best available evidence to unravel
the confounded effects of run sequence and S/V reveals a substantial in-
consistency in the S/V effect in two sets of aluminum runs.
Note, however, that the materials which exhibit the extremes as regards the
four principal components are Teflon and stainless with aluminum and Pyrex
in between. Since Teflon and stainless were adjacent in time with Pyrex
B-22
LOCKHEED MISSILES & SPACE. COMPANY. INC.
-------
LMSC-D406484
and aluminum both before and after, then it is clear that the possible drift
is not overwhelming material properties. Nothing more can really be said
because of the inconsistency between the early and late' and late aluminum
runs, without postulating a different kind of drift in the two places.
2. 8 Computer Program
Many times it is felt that a manual massage of experimental data presents
insight into the data that might be overlooked if the data were analyzed by
an impersonal computer. True as this may be, it is often the case that the
immensity of the analysis demands computer usage, and such was the
situation with many of the analyses of the smog chamber data. A computer
program, entitled the Modified Abbreviated Doolittle (MAD) program, was
therefore used in this study.
MAD is a generalized analysis program developed by the Brigham Young
University Department of Statistics, which is capable of analyzing balanced
and unbalanced univariate and multivariate analysis of variance and covariance
problems as well as univariate regression problems. These analyses can also
be performed in the presence of missing cells. The algorithm which is in-
corporated within the MAD program is an open ended algorithm allowing for
expansion to handle forms of analyses which can be based upon construction
of a X'X matrix. The X'X is then partitioned according to user control into
blocks containing the various analysis of variance terms and covariates, or
the independent variables with the dependent variables. The Abbreviated
Doolittle Operations are then performed on this particular ordering of the
X'X matrix, with those variables not specified by the user being ignored.
B-23
LOCKHEED MISSILES 8e SPACE COMPANY, INC.
-------
LMSC-D406484
3.0 COVARIANCE ANALYSIS
Fundamental to the study of the smog chamber data is a study of those
parameters which might be considered as covariates. Through such a study,
the desireability of including the covariates in a descriptive model can be
determined. Six parameters were considered as likely candidates for
inclusion as covariates: HCint, NO2> NOX> %NO2> TADJ*, and HC/NOX
These six parameters were studied separately and in various combinations
to determine their effect on the 23 dependent variables. The covariate effects
were determined for the dependent variables in both univariate and multi-
variate models(ALD Max and PANmax were excluded from the multivariate
study).
3. 1 Univariate Covariance Results
The results of the covariate study on the univariate variables indicated that
two covariates were generally useful in explaining data variation and that the
remaining covariates were not generally of additional value. The two useful
covariates in order of their relative importance were %NO? and HCint. The
adjustment coefficients for these two covariates are given in Table B3-1.
The data, as corrected by these factors, is given in Appendix A of this report.
Two approaches were used to determine the fact that %NC>2 and HCint were of
value, both of which invoked the usual analysis of covariance assumptions.
The first approach was to eliminate NO? and TADJ from the list of covariates
(on the basis of a preliminary investigation) and to test the effects of all
possible combinations of the remaining four covariates after accounting for
*TADJ represents that amount of time adjustment introduced in forming the
adjusted data set.
B-24
LOCKHEED MISSILES & SPACE COMPANY. INC.
-------
IMSC-D406484
TABLE B3-1
Univariate Covariate Adjustment Coefficient
Dependent
Variable
N00TM
HCT50
03TM
FWHM
XTBffi
1/03AVGR
1/HCAVGR
1/N02R
I/NO RATE
N02K)SE
N02DF
OoDOSE
HC Init
NO,
1/OoMAXR
1/HCMAXR
03MAXC
HCFC
PAWMAX
HCT75
HCT25
O^XT
AEDMAX
-18.
-17-013
-17.198
-1.0533
-3.2832
.067834
-.0258^3
-.018^21
-.0020981
-2.1619
-1.4307
-12.326
.19986
-.0090996
-.018591
-.15294
.12951
-.15136
2.4908
-12 .212
-7-3887
-10.464
-.34816
-2.2342
-2.4526
-2.2818
.23104
-2.4154
-.0054873
-.0014754
-.0023856
-.0025264
1.4o64
.098329
1.8671
.20911
-.00039398
-.000099554
.0046785
.0016781
.010004
-2.1539
-2 .44i6
.19681
.00025681
B-25
LOCKHEED MISSILES & SPACE COMPANY
-------
LMSC-D406484
a model of the four main effects (Materials, S/V, Spectrum, and cleaning)
and all but two of the two-way interactions. Using the reduction in sum of
squares approach to determine the effect of the covariate groups after the
model, the results were arrived at. The second approach was to use the
same model of main effects and interactions, and the reduction sum of squares,
but to allow all six covariates to enter in a forward selection procedure
analogous to the forward selection procedure used in Regression Analysis.
The results of the two approaches were in agreement in the univariate co-
variance studies.
3. 2 Multivariate Covariance Results
Using the two approaches described above, the dependent variables (excluding
ALD max and PAN max) were studied in a multivariate analysis to determine
the effect of the covariates. As opposed to the univariate results, three co-
variates were found to be useful in the multivariate description of the data.
The first approach (that of eliminating NO? and TADJ) resulted in the selection
of HCint, NO.., and %NO_ as jointly useful covariates, NO,, being added to the
-A. Li 1*.
univariate covariate set. The second approach, however, exchanged TADJ
for %NO9 to conclude the set should be HCint, NO , and TADJ. While the
dt j\.
two results appear to differ. such is not the case. Both sets are
effectively the same, since %NO2 and TADJ are highly correlated with a
coefficient of . 9579. Table B3-2 gives the adjustment factors associated with
the first set of covariates (HCint, NO,,, and %NO2), and Table B3-3 presents
the results of each step of the multivariate forward approach.
B-26
LOCKHEED MISSILES & SPACE COMPANY. INC.
-------
IMSC:-D4o64&4
TABLE B3-2
MUI/TIVARIATE COVARIATE ADJUSTMENT COEFFICIENTS
Dependent
Variable EC Inlt jp % N0r
NO TM -47.281 -9-975 -2.170
HCT50 -46.155 4.019 -2.462
03TM -75.241 5.024 -2.848
FWHM -38.021 8.494 -.2197
XTIME -11.965 -8.251 -2.270
1/03AVGR -.0905 -.0091 -.0074
1/HCAVGR -.0563 -.0114 -.0015
1/N02R -.0225 . u.o68l -.0015
1/NDRATE -.0355 -.0658 -.0019
N02DOSE -71-297 118.25 -.4304
ND2DF -16.073 1.495 -.0683
03DOSE 23.464 18.987 1.881
O^OF 6.189 -20.859 .4407
1/03MAXR -.0264 .0126 -.0009
1/HCMAXR -.0295 -.0034 -.0003
03MAXC -.0955 -.0616 .0063
HCFC -.0710 -.0606 -.0025
HCT75 -19.472 -4.067 -1.998
HCT25 -47.829 -28.033 -2.474
N02XT -33.787 12.999 -.0869
03XT -61.863 20.286 -.6164
B-37
LOCKHEED MISSILES & SPACE COMPANY
-------
LMSC-D406484
TABLE B3-3
Results of Covariate Multivariate Forward Approach
Covariate Alone " .
Covariate HC Init BO N0p % NO Tadj HC/NOX DF F = .05
""
X
p X
^ ^
5.982 37-797 8.015 10.385 13-268 7-45^ 21,15 2.33
Covariate After NOV
J\.
5.566 - 5.654 8.274 12.247 3-667 21, 14 2.39
Covariate After NO and Tadj
"F" 5-139 - 0.894 0.865 - 3.789 21,13 2.46
Covariate After NO , Tadj, and HC Init
0.806 0.809 - 1.438 21,12 2.54
Covariate After NOX, Nadj, HC Init, HC/NO , and NO
0.789 - - - 21,10 2.77
B-28
LOCKHEED MISSILES & SPACE COMPANY
-------
LMSC-D406484
3. 3 Comparison of Covariate Corrected Data to Adjusted Data
Upon completion of the experiment, the experimenter adjusted the data to a
common basis of initial conditions. The analyses reported in this present work
have been performed on the original data prior to the experimenter's adjust-
ment. An effort has been made to compare this original data corrected by
covariates to the experimenter's adjusted data. To do this, both sets of
data have been univariately analyzed according to the same model, and the
2 2
R values therefrom compared. The R values thus attained for each of the
23 dependent variables are represented in Table B3-4 along with the sign of
the difference. The data with the higher R values is the preferred data as
the higher R indicates that more of the variation has been accounted for by
the adjustment. In all but one case, the covariate adjustment is superior to
the experimenter's adjustment with the one exception being, understandably,
Ozone DF. As such, the analyses were continued using this covariate
corrected data. The experimenters adjustment, based solely on %NO~ was
typically 1 or 2 percent lower in R .
4. 0 ANALYSIS OF VARIANCE
Having determined the significant covariates for the univariate and multi-
variate models, attention was focused on the models themselves. The basic
model investigated was Y = A* t *> /?X. .. . 4- M. + A. + B. + C,
" ijklm / ^~l ijklm i j k 1
MA..+MB.. + AB., + MAB...
ij ik • jk ijk
+ MC., + AC.. + £.., ,
il jl *— ijklm
where ff was an adjustment factor for the appropriate covariate X, M was
materials (i = 1, . . . , 5, with original and rerun aluminum considered as two
materials), A was S/V (j = 1, 2), B was Spectrum (k = 1, 2), and C was
B-29
LOCKHEED MISSILES & SPACE COMPANY. INC.
-------
TABLE B3-4
Comparison of Covariate Corrected Data to Adjusted Data
Dependent
Variable R - Covariate Corrected Data R
NOpTM
HCT50
0_TM
FWHM
XTIME
1/0 AVGR
1/HCAVGR
1/N02R
1/NORATE
N02DOSE
N02DF
0 DOSE
0 DF
1/O.MAXR
1/HCMAXR
0 MAXC
HCFC
PANMAX
HCT75
HCT25
N02XT
0 XT
AIJDMAX
95-9
96.3
95-7
93-7
96.2
92.1
95-3
9^.6
95-3
88.3
90.0
93-6
91-9
82.5
92.8
92.8
78.9
87.1
. 93 A
96.5
91-9
93-2
77-8
(96.1)*
(95.9)
(93-9)
(96.3)
(93-5)
(95-9)
(95.6)
(96.3)
(90.6)
(9Q.6)
(95.Q)
(85-3)
(93-3)
(95.2)
(82.6)
(88.5)
(93-9)
(96.8)
(92.1)
(93.6)
(80.7)
- Adjusted Data
94.9
95-5
92.7
95-8
91.6
93-2
93-8
87.3
88.5
92.6
92.8
79-9
90.9
91-3
78.3
82.6
92.9
95-5
90.3
91.0
7^.2
IVDDEL: Y =
2
*( ) = R with all six covariates considered
B-30
Diff
LOCKHEED MISSILES & SPACE COMPANY
-------
LMSC-D406484
cleaning (1=1, 2). All main effects were considered fixed, with contrasts
used given in Table B4-1. This model was analyzed in two ways: first as
a univariate model for each of the 23 dependent variables, and secondly, as
a 21 variable multivariate model.
TABLE B4-1
CONTRASTS USED IN THE ANALYSES OF VARIANCE
Contrast
Material
Original Stainless Rerun
Aluminum Pyrex Teflon Steel Aluminum
1
2
3
4
Contrast
S/V
-1
-1
0
-1
Condition
S/V Low S/
-1
-1 -1 4
000
-1 1 0
1 1 0
V High
1
-1
1
0
-1
Cut Full
Spectrum -1 1
Purge Clean Vacuum Clean
Clean -1 1
n
Interactions formed by appropriate products of above.
B-31
LOCKHEED MISSILES & SPACE COMPANY. INC.
-------
LMSC-D406484
4. 1 Results of Univariate Analysis
The univariate analysis resulted in a preponderance of material, spectrum,
material by S/V, and material by spectrum effects. These four terms were
significant for at least 22 of the 23 dependent variables. The three-way inter-
action of material x S/V x spectrum was surprisingly more prevalent than
either the S/V or cleaning main effects, though none of these effects appeared
in even half of the cases. Table B4-2 shows those terms which were of
statistical significance for each dependent variable.
Three different considerations were made in determining significance. These
considerations were necessitated by the imbalancing of the data, which resulted
in nonorthogonal sums of squares. The first consideration was to test each
term against the error in the order specified by the model of Section 4. 0.
The second consideration was to test each term using the method of weighted
squares of means. Thirdly, each term was tested after adjusting only for the
mean. In each consideration, each term was tested against the error mean
square, which remained constant for all three considerations. The results
of these three approaches were then compared, with the final results given
above and in Table B4-2. Prediction coefficients for the complete model are
given in Table B4-3. These prediction coefficients apply to the data after it
has been adjusted for HCint and %NOy-
The material effect and material interactions each were composed of four
degrees of freedom. As each degree of freedom was defined by a material
contrast (see Table B4-1), these contrasts were further investigated. Since
there were no missing cells and due to the basic orthogonality of the contrasts,
each degree of freedom is directly interpretable from the contrasts. Hence,
B-32
LOCKHEED MISSILES & SPACE COMPANY. INC.
-------
TABLE
Significance of Effects & Interactions on Univariate Data Given HC Init and $ N00
2
BC
N02TM
HCT50
03TM
FWHM
XTIME
1/0 ANGR
1/HCANGR
I/NO R
cL
I/NO RATE
N02 DOSE
N0_ DF
OJXJSE
3
0^ DF
3
1/0_MAXR
3
1/HCMAXR
0-MAXC
3
HCFC
PANMAX
HCT75
HCT25
NOgXT
0_XT
3
ALDMAX
M A
X
X
X
X
X
X X
X
X
X
X
X
X X
X X
X X
X X
X X
X
X X
X
X
X
X
X
B
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
MA
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
MB AB MAB C MC AC
X XX
X X
X XX
XXX
X X
X X
X
X XXX
X X
X
XXX
X X
X X
X XX
X
XX XX
X ' X
XXX XX
X X
X
X X
X X
X X
Total 23 7 22 22 23 5 9 86
B-33
LOCKHEED MISSILES & SPACE COMPANY
X
-------
TABLE B4-3
Univariate Prediction Coefficients for Covariate Adjusted Data
Term
CONSTANT
MAT 1
1- MAT 2
OMAT 3
0 MAT 4
IS/7
U| SPEC
0 MAT 1
, MAT 2
55 MAT 3
(/) MAT 4
F MAT 1
!5 MAT 2
MAT 3
* MAT 4
iJS/Vx
> MAT 1
jjj MAT 2
MAT 3
g MAT 4
x S/V
x S/V
x S/V
x S/V
x SPEC
x SPEC
x SPEC
x SPEC
SPEC
x S/V x
x S/V x
x S/V x
x S/V x
2 CLEANING
£ MAT 1
Z MAT 2
•< MAT 3
MAT 4
S/V x
x CL
x CL
x CL
x CL
CL
SPEC x CL
NOgTM
107
-10
2
.10
18
_
-23
-2
3
-15
4
2
-1
-8
' -1
SP
SP
SP 6
SP -3
4
1
5
-1
• 29
• 311
.752
.410
• 791
• 5319
.809
.214-2
.368
.542
.013
.81*3
.850
• 7315
.881
.362
.1935
.0341
• 530
.900
.799
.153
.9111
.283
.6221
.2663
.042
HCT50
133.64
-10.172
2.109
10. ?ln
17.187
.3122
-26.266
-2.322
5-995
15.962
5-893
2.436
•-.8o47
1.144
-8.643
.3043
.4254
.0588
4.941
-3.699
3-287
1.265
2.075
4.346
-1.666
-.9434
-1.082
0_TM
170.96
-9.654
9.148
19-481
21.960
1.407
-32.929
-2.238
4.6l8
-23.300
4.209
2.971
-3-157
1.441
^10.813
-2.286
.3660
.1241
12.910
-3.034
5.673
1.458
-.799
4.088
-.7803
-1.170
-1.668
FWHM
100.07
-3.459
7.055
11.019
10.278
.1149
XTBffi
55.158
-6.440
.2316
3-435
10.225
-.2989
-22.772 -10.649
-1.197
2.733
-11.662
.4843
1.855
-.5144
-2.877
-7.958
-.4999
.0459
-.9418
9.442
-1.054
.1613
1-376
-2.004
.5853
-3.159
-.0716
-1.109
-1.408
1.296
-3.453
2.551
1-197
-.4773
1.884
-3.632
.0514
.0725
.4588
1.496
-1.621
3.530
.5363
1.026
2.479
1.623
.1823
-1.104
1/0 AVGR
.2690
-.0220
.0008
.0178
.0396
.0151
-.0793
-.0049
.0049
-.0483
.0097
.0065
.0078
-.0045
-.0277
-.0081
.0025
-.0005
.0352
-.0078
-.0081
.0067
-.0041
.0071
-.0123
-.0035
.0027
l/HCAVGR
.0934
-.0073
.0030
.0089
.0138
-.0003
-.0197
-.0017
.0052
-.0120
.0055
.0020
.0011
0.0
-.0080
-.0002
.0006
-.0011
.0055
-.0035
.0012
.0013
.0014
.0039
-.0017
-.0016
-.0003
1/N02R
.0854
-.0106
-.oo4i
.0043
.0138
-.0013
-.0192
-.0021
.0004
-.0025
.0039 .
.0026
.0029
.0041
-.0041
.0010
-.0001
.0002
.0009
-.0031
.0054
.0012
.0022
.0060
.0028
.0002
-.0022
1/NORATE
.1089
-.0100
.0096
.0065
.0136
.0001
-.0214
-.0020
.0059
-.0136
.0051
.0021
-.0009
.0047
-.OOf 2
-.0003
.0006
.0001
.0064
-.0046
.0009
.0012
-.0002
.0034
-.0008
-:.OOO4
-.0025
td
CO
rfk
-------
TABLE B4-3 (Continued)
o
o
X
m
m
O
2
U)
f A
u)
m
(/)
a>
U)
5
o
m
0
0
2
dk
TJ
3>
-<
Term
CONSTANT
MAT 1
MAT 2.
MAT 3
MAT 4
S/V
SPECTRUM
MAT 1 x S/V
MAT 2 x S/V
MAT 3 x S/V
MAT 1* x S/V
MAT 1 x SPEC
MAT 2 x SPEC
MAT 3 x SPEC
MAT 4 x SPEC
S/V x SPEC
MAT 1 x S/V
x SPEC
MAT 2 x S/V
x SPEC
MAT 3 x S/V
x SPEC
MAT 4 x S/V
x SPEC
CLEANING
MAT 1 x CLEAN
MAT 2 x CLEAN
MAT 3 x CLEAN
MAT 4 x CLEAN
S/V x CLEAN
SPEC x CLEAN
N02
175
-5
14
19
IT
3
-22
-1
6
-31
3
-5
_
-13
-4
,
-l
14
. -1
1
6
-4
-
DOSE
•72
.107
.542
.965
• 799
.052
• 917
• 510
.860
.924
.850
.9097
.709
.4611
.221
.893
.7581
.027
.899
.363
,8486
.528
.4972
.048
.860
.1190
.2531
N02DF
39.167
-1.138
2.430
4.566
3-625
.4806
-5 -413
-.2272
1.484
-6.797
• 7314
.3246
-1.072
.3276
-2.971
-.9605
.0893
-.6885
3.6650
-.4994
.0362
.4799
.0615
1.146
-1.272
-.2251
-.1840
0 DOSE
144.18
4.281
.4174
.0629
-10.626
-11.578
27.926
-.6670
-.4817
14.965
-.6073
-1.210
-.5213
6.632
9.404
-4.099
-.6160
1.378
-8.340
3-559
3.678
-1.073
1.076
-3.991
3-917
-.4989
• 3944
0 DF
32.120
.8797
-1.020
.2885
-2.824
-2.717
5.852
-.1251
-.2603
3-740
-.1963
-.1188
.2679
1.931
2.263
-.9143
-.1249
-.0915
-1.293
.4434
. .4211
-.2530
.0895
-.6963
.4579
-.6088
-.2772
1/0 MAXR
.0469
.0012
.0006
.0025
.0029
.0064
-.0100
.0011
.0016
-.0073
-.0011
-.0007
.0002
-.0020
-.0034
-.0005
-.0005
.0013
.0070
-.0001
-.0022
.0003
-.0011
.0005
-.0030
.0006
.0021
1/HCMAXR
.0513
-.0018
.0024
.0093
.0066
.0024
-.0108
-.0005
.0009
-.0115
-.0015
.0007
-.0015
-.0019
-.0053
-.0013
.0003
-.ooo4
.0029
.0003
-.0012
0.0
.0003
-.0026
-.0012
.0002
.0005
0 MAXC
.9678
-.0111
.0358
.04 12
.0055
-.0680
.0718
-.0135
-.0081
.0170
.0083
.0071
-.0287
.0222
.0131
-.0116
-.oo4i
.0128
-.0146
-.0027
.0426
-.0033
.0255
.0009
.0306
-.0071
-.0057
HCFC
.2037
-.0136
-.0153
.0078
.0443
.0010
-.0776
-.0052
.0267
-.0572
.0102
.0105
.'0223
-.0154
-.0259
-.0061
.0067
.oo4o
.0321
.0011
-.0200
.0051
.0130
.0290
-.0095
-.0213
.0158
PANMAX
.2750
-.0054
.0158
.0708
-.0666
.0248
.0453
-.0113
-.1136
-.0338
-.0504
-.0093
- . 1169
.0354
-.0061
.0273
.0003
-.0349
-.0242
-.0518
-.0038
.0047
.0661
.0391
.0405
-.0214
-.0054
o
w
L
01
-------
TABLE B4-3 (Continued)
Term
CONSTANT
O
o
MAT
MAT
MAT
MAT
S/V
1
2
3
4
£ SPECTRUM
m
m
2
(A
(A
m
(A
fi>
(A
TJ
O
m
0
O
2
T)
2
MAT
MAT
MAT
MAT
MAT
MAT
MAT
MAT
S/V
MAT
MAT
MAT
MAT
1
2
3
4
1
2
3
4
X
1
2
3
4
x S/V
x S/V
x S/V
x S/V
x SPEC
x SPEC
x SPEC
x SPEC
SPEC
x S/V x
x S/V x
x S/V x
x S/V x
SPEC
SPEC
SPEC
SPEC
CLEANING
MAT
MAT
MAT
MAT
S/V
SPEC
1
2
3
4
X
x CLEAN
x CLEAN
x CLEAN
x CLEAN
CLEAN
x CLEAN
HCT75
93
-8
-
4
12
-1
•17
-1
6
-7
7
2
2
4
-4
_
3
-4
3
1
2
4
_
-2
-
.168
.270
.7944
.224
.380
.891
.610
.798
.091
.265
• 73^
.114
.034
.094
.088
.2231
.2173
.1019
.105
.494
.581
.478
.187
.062
.4469
.021
.7448.
HCT25 NO-XT
169
-9
13
18
1
-31
-2
6
-18
5
2
1
-8
-
7
-3
1
1
2
4
-1
-2
-
..10
•909
.9236
.186
.641
.676
.478
.129
.317
.650
.333
.577
.165
.1084
.875
.9889
• 37^0
.2718
.273
.576
.650
• 552
.532
• 555
.082
.035
•0371
51
-3
2
6
8
_
-12
-
2
-12
1
-1
-4
-1
_
6
-1
1
_
4
_
-
•915
.849
.462
.172
• 079
.8581
.955
.7359
.162
.559
.8928
• 521
.096
.2755
• 578
.132
.0921
.4210
.294
.652
.724
• 5325
.1801
.180
• 3^19
.3865
.8231
0 XT
115
-3
9
16
11
l
-22
_
3
:-l9
1
1
-2
-
-7
-2
_
11
-1
1
-1
1
-2
-1
0
•98
.188
.290
.111
.821
.530
.397
.7426
.527
•^37
.664
.893
.475
.3139
.147
• 5^3
.4048
.4622
.389
.291
• 857
.9005
.861
.562
.268
.552
• 7057
ALDMAX
m
.
m
•
•
"• •
•
*
_ m
•
~ *
™ •
•" •
•• •
•
— •
~ •
•
•
•
•
7396
0083
1496
0092
0006
0049
0126
0010
0632
0047
0070
0016
0036
0209
0399
0235
0021
0079
0071
0133
0031
-.0099
— .
_ .
.
_ .
.
0129
0142
0004
0022
0188
w
CO
Oi
-------
LMSC-D406484
the first material degree of freedom compares stainless steel with the
average of the other materials. The second degree of freedom compares
the original aluminum to the redone aluminum, the third, pyrex to teflon,
and the fourth, the average aluminum to the average of the pyrex and teflon.
The material interactions follow the basic pattern. The significance of these
individual degrees of freedom is shown in Table B4-4. Of particular concern
was the comparison of the aluminums, wherein 8 of the 23 variables indicated
a significant difference.
With such profound material effects and interactions are shown in Table B4-4,
it was thought that a univariate analysis by materials might be of value, as
the various terms could then be considered separately from the material
interactions. A study was thus performed which, after adjusting for the co-
variates HCint and %NO?, checked the need of additional covariates and then
studied the model of Y... . = JU + ^AX... . + A. B. + AB.. + C. + AC..
ijkl 1^ ^ P ijkl i j ij k ik
with the same term explanations as in Section 4. 0.
The original aluminum data, studied in this new context, required no additional
covariates after HCinit and %NO2. Spectrum, S/V, and spectrum by cleaning
were generally significant, though no consistent pattern in the model was
detected.
The Pyrex analysis helped explain the need for NO as a covariate in the multi-
variate analysis (see Section 3. 2). Though the error had only 1 degree of
freedom, NO,, was significant after HCinit and %NO? nearly 20% of the time.
The model, however, was analyzed ignoring this NO,., covariate, with the result
of S/V and spectrum being the predominately significant terms.
B-37
LOCKHEED MISSILES & SPACE COMPANY. INC.
-------
TABLE
Significance of Univariate Material Contrasts Given HC Init and % NO,.
o
o
*
X
m
m
p
2
U)
en
r
m
en
R>
en
2
^
o
m
O
0
2
D
>
Z
<
NOpTM
HCT50
0 TM
\s o
FWHM
XTIME
1/03AVGR
1/HCAVGR
1/N02R
1/NORATE
w N02DOSE
i, N02DF
». 03DOSE
03DF
1/03MAXR
1/HCMAXR
03MAXC
HCFC
PAN MAX
HCT75
HCT25
vrf\ "vm
JNOpAl •
03XT
ALDMAX
MAT
1
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
2
X
X
X
X
X
X
X
X
3
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
k
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
1
X
X
X
X
X
X
X
X
X
X
X
MAT x 8,
2
X
X
X
X
X
X
X
X
3
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
k
X
X
X
X
X
X
X
X
X
X
TOTAL
Overall
Term Significant
21 8 15 20
23
11 8 20 10
22
MAT
1 2
X
X
X
X
X
X
X
X
X
X
X
X X
X
X
X
X
X
X X
x SPEC
3 k
X
X
X
X
X
X
X
X
X
X
X
X X
X X
X
X
X
X
X
X
X
X
16 k 2 21
23
MAT x S/V x SPEC
1 2 3 ^
X
X
X
X X
X
X
1 1 11 9
9
MAT x CLEAN
1234
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X X
X X
X
X
X X
X
X
X X
XXX
o
X Indicates Significance
-------
LMSC-D406484
The Teflon data required no new covariates after HCinit and %NO_.
C*
Spectrum, S/V, and the S/V by spectrum interaction were the most
significant, with an occasional cleaning effect and cleaning effect inter-
action.
Cleaning appeared as a significant term in the stainless steel analysis, along
with S/V and spectrum. No additional covariates were required.
The rerun aluminum differed from the original in that only spectrum was
significant in the rerun aluminum.
A major difficulty in this 'by material" study was the lack of error degrees
of freedom, thus a limited power for detecting significance. Hence a
failure to declare a term significant in no way implies an absolute lack of
significance. With this understanding, Tables B4-5 to B4-9 summarize the
above "by materials" analyses.
4. 2 Results of Multivariate Analysis
The multivariate data, adjusted for HC Init, TADJ, and NO,, showed most
of the terms of the model of Section 4 to be significant, when the three methods
of Section 4. 1 were considered. Only S/V by spectrum, S/V by cleaning, and
spectrum by cleaning failed to be significant at o< = . 05. Table B4-10 presents
an abbreviated MANOVA in which the numerator and denominator degrees of
freedom and associated "F" is given for each term, where the "F" is cal-
culated from Wilk's U. (Analogous univariate tables were deemed impractical
for the analyses of Section 4. 1, due to the large number of such tables which
would be required). Though a multivariate "by materials" study would be of
great interest, such a study was not possible due to limited error degrees of
freedom.
B-39
V-'
LOCKHEED MISSILES & SPACE COMPANY. INC.
-------
TABLE BU-5
Univariate Original Aluminum Analysis, After HG Init and % N0p
S/V SPEC S/VxSPEC CLEAN S/V x CLEAN SPECxCLEAN
HCT50
0 TM
FWHM
XTIME
1/0 AVGR
1/HCAVGR
1/N02R
1/NORATE
N02DOSE
NO DF
O^DOSE
0 DF
1/HCMAXR
0-MAXC
HCFC
PAN MAX
HCT75
HCT25
°3XT
ALDMAX
TOTAL
NO
0
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X X
X
X X
X
X
X X
X
X
X
X
X
X
X X
X
X X
X X X
X X
X X
X X
X X
X
X
X
X
X
X
X
16
21
X Indicates Significant Term
X
X
X
X
X
X
X
B-46-
LOCKHEED MISSILES & SPACE COMPANY
-------
TABLE B^-6
Univariate Pyrex Analysis After H C Init and
NO S/V SPEC S/V x SPEC CLEAN S/V x CLEAN SPEC x CLEAN
N02TM
HCT50
0-TM
FWHM
XTIME
1/0 AVGR
1/HCAVGR
1/N02R
1/NORATE
N02DOSE
NOgDF
0-DOSE
3
0_DF
3
1/0 MAXR
1/HCMAXR
O.MAXC
3
HCFC
PAN MAX
HCT 75
HCT 25
N02XT
0-XT
X
X
X
X
X
X
X
X
X X
X X
X X
X
X
X
X X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
TOTAL 4 21 22
X
X
X
X
X
B-41
LOCKHEED MISSILES & SPACE COMPANY
-------
TABLE
Univariate Teflon Analysis, After HS Init and % HK)f
NO
S/V x SPEC CLEAN S/V .x CLEAN SPEC x CLEAN
N02TM
HCT50
0 TM
FWHM
XTIME
1/0 AVGR
1/HCAVGR
1/N02R
1/NORATE
ND2DOSE
N02DF
0-DOSE
0 DF
1/0-MAXR
1/HCMAXR
0 MAXC
HCFC
PAN MAX
HCT 75
HOT 25
NO XT
0 ST
ALDMAX
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
TOTAL
22
•8
X Indicates Significance
B-42
LOCKHEED MISSILES & SPACE COMPANY
-------
TABLE
Univariate Stainless Steel Analysis, After HC Init and % NO,.
NO,
S/V x SPEC CLEAN S/V x CLEAN SPEC x CLEAN
NOJTM
HCT 50
0-TM
3
FWHM
XTIME
1/0 AVGR X
1/HC AVGR
1/HD2R
1/NORATE
N02DOSE
N02DF
OJXJSE
3
°3DF
1/0.,MAXR
1/HC MAXR
0 MAXC
HCFC
PAN MAX
HCT 75
HCT 25
N02XT
0-XT
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X X
X X
X
X
X X
X
X
X
X
X
X
X
X
X
X
X
.
X
X
X
X
X
X
X X
X X
X
X
X
X
X
X
X
X
X
X
ALDMAX
TOTAL
18
21
15
X Indicates Significance
B-43
LOCKHEED MISSILES & SPACE COMPANY
-------
TABLE
Univariate Rerun Aluminum Analysis, After HC Init and $ N0r
HCT50
0 TM
FWHM
XTIME
1/0-AVGR
1/HCAVGR
1/N02R
1/NORATE
0 DOSE
0 DF
1/0 MAXR
1/HMAXR
0 MAXC
HCFC
PANMAX
HCT 75
HCT 25
0 XT
ALDMAX
X
1/V
X
X
X
X
X
X
X
X
SPEC S/V x SPEC
X
X
X
X
X
X
X
X
X
X
X X
X
X X
X
X
X
X
X
X
X
X
CLEAN S/V x CLEAN SPEC x CLEAN
X
TOTAL
8
20
0
0
X Indicates Significance
B-44
LOCKHEED MISSILES & SPACE COMPANY
-------
TABLE Bk-10
Abbreviated MANOVA, Given HC Init, TABJ, and NO
Term
MATERIALS
S/V
SPECTRUM
MAT x S/V
MAT x SPEC
S/V xSPEC
MAT x S/V x SPEC
CLEAN
MAT x CLEAN
S/V x CLEAN
SPEC x CLEAN
Numerator df
&
21
21
8^.
&
21
&
21
&
21
31
Denominator
30
7
7
30
30
T
30
7
.3
7
7
9.097**
13.836**
50.200**
U.993**
3.320**
2.003
3-895*
1.800*
1.395
*Indicates Significance at a= 0.5
**Indicates Significance at a. = .01
B-45
LOCKHEED MISSILES & SPACE COMPANY
-------
LMSC-D406484
5. 0 PRINCIPAL COMPONENTS
5. 1 Selection
Tables B5-1 to B5-3 present the correlation matrix of the 23 parameters
derived from the raw data, the adjusted data and the data after the covariate
correction (for HC init and %NO2) respectively. In each case the order of the
parameters has been organized so that highly correlated parameters are
adjacent to each other.
Rather than using the reported values of the rate parameters, the inverse
values have been used. This was done because it was noted that the correlations
between these rates and the time parameters increased when this transformation
was made. The cross plots show a non-linear relation as typified by Fig.
B5-1.
Generally, all three correlation matrices are similar although there is a
slight tendency for the values in Table B5-3 to be higher.
There is a remarkable regularity among the first 13 parameters - they are
very highly correlated with each other. So much so that we are led to believe
that to a major extent they are all reflective of a single underlying "TIME"
parameter of the system irregardless of materials, S/V, spectrum or cleaning.
There is a similar linkage between parameters 8 and 19 and between parameters
3 and 18. This is as would be expected since these are in each case standardized.
versions of one another.
Also all four (3/18 and 8/19) exhibit similar correlations with the 13 time
parameters.
B-46
LOCKHEED MISSILES & SPACE COMPANY, INC.
-------
2 21 io 11 12. 7 20. 23 22 • 5 i4 IT 1 8 19 3 " 18 4 13 6 • 9 16
2
21 96l
10 957 937
11 991 944 968
12 978 921 947 992
7 973 912 913 976 979
£ 20 901 792 824 919 938 955
ft 23 876 750 798 893 915 954 968
* 22 9^7 835 879 9^ 9^ 951 934 932
S 5 921 843 869 938 959 9^5 9^5 9l6 915
tn 14 983 924 961 992 987 969 920 895 944 939
0 17 968 942 935 970 959 953 885 858 904 916 961
Z 1 939 963 934 926 907 868 771 704 830 8l4 9l4 913
en s -750 -670 -686 -783-826 -800 -853 -808 -763 -900 -778 -754 -652
^ 19 -753 -670 -684-782 -815 -801 '-849 -8ll -772 -890 -772 -744 -635 974
£ 3 84i 695 761 853 855 895 901 9^5 920 870 859 812 632-729 -766
w 18 863 720 786 877 887 917 925 961 931 898 885 844 670-761 -769 986
g> 4 4,16 287 324 463 538 556 674 686 508 681 470 421 224-792 -781 622 643
w 13 808 696 675 820 855 867 899 905 853 864 824 791 655 -791 -787 864 889 665
T> 6 037 046 -oil -103 -185 -139 -293 -261 -109 -372 -105 -079 052 655 621-191 -2i4 -777 -294
£ 9 774 666 748 805 84i 790 824 792 824 881 822 749 674-810 -791 654 780 638 756-391
m 16 -161 -185 -226 -177 -199 -184 -197 -163 -11^ -213 -189 -226 -l4o 171 176 -081 -087 -226 -080 221 -260
o
o
•s.
> . TABLE E5-1
-<
Correlation Coefficients for Unadjusted Data
W
-------
Table B5-2
CORRELATION COEFFICIENT FOR ADJUSTED DATA
10
11
7 22 20 23 21 5 14 1 17 8 19 3 18 4 13 6 9
0
o
*
I
m
m
o
(ft
en
m
(ft
(ft
TJ
>
O
m
8
TJ
Z
n
w
I
oo
2
10
11
12
7
22
20
23
21
5.
1
17
8
19
'3
18
4
13
6
9
2
10
11
12
7
22
20
23
21
5
4
1
17
8
19
3
18
4
13
6
9
•953
•991
•977
•971
•971
.895
.863
971
867
916
•767
829
-7l4
-732
847
^
800
016
785
2
924
989
946
955
379
838
964
942
902
881
.800
768
951
843
922
784
820
-642
-656
754
766
263
621
.022
745
10
863
959
9*0
914
266
642
991
974
956
•911
880
955
891
929
747
847
-746
-762
86l
873
409
762
-079
814
11
TIME
940
993
929
962
423
802
979
960
•938
910
925
900
912
705
829
-785
-795
878
894
478
800
-l!J4
850
12
956
989
899
949
500
838
971
•958
953
910
874
875
672
804
-764
-782
914
926
505
811
-126
814
7
950
967
869
948
531
864
934
932
787
830
844
647
773
-757
-779
923
929
460
780
-119
847
22
941
953
837
920
514
856
968
798
846
789
551
742
-842
-848
907
926
642
84o
-306
846
20
957
910
755
889
680
90S
745
798
745
477
704
-783
-800
952
963
650
858
-276
816
23
923
882
687
860
694
909
844
919
830
817
-627
-644
726
733
223
630
091
686
21
844
938
986
923
222
681
918
767
880
-754
-781
-770
-776
523
764
-277
-694
5
942
799
912
-881
-884
907
920
658
859
-354
891
892
907
-568
-782
-735
-740
269
689
-060
-662
14
916
950
-746
-756
862
878
438
806
-809
839
781
-306
-345
-498
-496
056
4i4
-319
-4l4
1
1
RATE
896
-600
-599
645
668
160
629
.121
652
-534
-563
-693
-706
312
681
.020
-588
17
-715
-722
830
853
405
787
-094
762
974
-718
-743
-783
702
663
-819
8
-777
-756
-732
-764
-782
-723
637
19
-772
-769
987
560 574
826 834 625
-227 -242 805 400
802 815 -557 -673 -368
18
13
O
648 662
886 902 664
-777 -290
636 762
o
OS
#t
-------
2 10 ll 12 7 22 20 23 21 5 14 . l 17 8 19 3 18 4 -13 6 9 16
2
10 951
11 989 964
12 976 94i 990
7 969 899 970 975
22 958 880 954 957 972
20 888 800 904 930 955 934
523 859 765 873 901 952 933 966
O21 967 945 950 921 899 868 780 726
* 5 920 866 938 960 942 932 94i 905 839
ml1*- 983 957 993 988 969 955 9l4 883 933 948
N l 933 934 921 899 852 823 740 664 979 811 903
017 966 931 969 957 947 924 878 843 937 911 965 910
28 -752 -678 -786 -830 -800 -790 -856 -807 -655 .891 -794 -652 -747
w 19 .711.9 _66i -776 -8ll -800 -788 -848 -811 -651 -871 -781 -621 -731 970
£3 842 746 851 860 913 917 906 953 706 875 865 611 820 -742 -776
£ 18 860 770 872 887 927 930 925 96l 723 901 887 648 847 -770 -774 986
w 4 375 667 420 500 525 508 662 677 208 638 439 169 368 -771 -774 631 64o
* 13 . 785. 633 794 835 858 848 892 898 666 850 802 608 769 -799 -799 871 887 667.
y, 6 -047 -019 -117 -201 -146 -i4i -312 -276 065 -371 -135 -032 -078 656 619 -208 -233 -785 -326
2 9 7&0 750 809 852 805 829 830 796 681 890 838 671 762 -810 -777 754 784 619 754 -393
016 -137 -203 -154 -176 -155 -139 -202 -166 -120 -186 -179 -122 -130 -125 -139 -105 -098 -183 -070 186 -280
m 15 -156 -247 -191 -174 -.061 -098 -018 046 -200 -135 -179 -268 -108 135 065 022 -035 o4o -ooo 093 -178 -002
O
o
2 TABLE B5-3
Correction Coefficients for Covariate Adjusted Data
k
o
S>
W+-
5-
(O
-------
210
190
170
150
CM
Z 130
CN
1
110
90
70
50
30
A = ALUMINUM
P = PYREX
T = TEFLON
S = STAINLESS STEEL
PTT
LMSC-D406484
Rg. B5-1
10
20 30
PARAMETER 1 (NO2R)
B-50
40
50
-------
LMSC-D406484
Finally it is notable that parameters 15 and 16 are largely uncorrelated with
each other and with any other parameters.
Thus, we are led to expect that the full 23 parameter set might be well
approximated by a reduced set involving:
1 or more components representing "TIME"
1 or more components representing "DOSE"
1 or more components representing the set consisting
of 4, 13, 6 and 9 plus possibility parameters 15 and 16.
Prior to forming the principal components it was necessary to consider how the
variables should be scaled with respect to each other. To this end, the total
variances across the full set were compared parameter by parameter against
the variances of replicate observations (prior to covariate adjustment). These
ratios shown in Table B5-4, were quite similar with the exception of parameter
15, where the ratio was quite low - indicating that, since the error variance
was comparable to the total variance, there was no point in considering this
parameter further. Thus parameter 15 was dropped from further consideration.
It is reassuring to see the relatively uniform value of this ratio for the other
parameters, since it is indicative of good experimental practice.
5. 2 Calculation
In forming the principal components, each parameter value was divided by its
"error" standard deviation and the principal components were extracted
from covariance matrices of these standardized values.
B-51
LOCKHEED MISSILES & SPACE COMPANY. INC.
-------
TABLE B5-4
Ratio of Mean Variation to Error Variation
LMSC-D4o64'84
PARAMETER
2
10
11
12
7
20
21
(22)
(23)
5
Ik
1
17
3
18
8
9
4
13
6
9
16
15
T (OVERALL)
TIME (ERROR)
1*2.896
33.149
43.650
^6.967
52.069
29.649
23.168
20.764
32.338
Inverse Avg Rate
.1140
.0343
.0372
.0394
DOSE
42.943
9.480
37-146
8.269
Inverse Max Rate
• 0159
.0174
.13^0
.1120
7-35
8.3U
6.84
6.76
8.05
.4.83
4.18
2.91
4.86
.01149
.00623
.00914
.01043
8.43
1.93
8.51
1.68
& Concentrations
.00567
.00565
.033
.039
.194
.135
PAN MAX
AID MAX
.039
• 135
RATIO
5.836
3.372
6.468
6.139
5-543
9.922*
5.506
4.070
3.778
5.094
4.912
4.365
4.922
2.8o4
3.080
4.061
2.872
4.974
1.000
B-52
LOCKHEED MISSILES Sc SPACE COMPANY
-------
LMSC-D406484
5. 2. 1 Time
In the case of the 13 time parameters -
1, 2, 5, 7, 10, 11, 12, 14, 17, 20, 21, 22, 23
we have a certain amount of redundancy. For example parameters 22 and
23 are linear combination of others in the set. Also, parameters 10, 11 and
12 are really three related readings from the same (HC) curve. Hence, it
was decided to use a reduced set to derive the principal components. This
set consisted of parameters:
1, 2, 5, 7, 12, 14, 17, 20, and 21
The linear combination of these variables which has maximum variance has
the following coefficients respectively:
.32, .34, .33, .34, .34, .34, .34, .31, .32
This linear combination accounts for 92. 6% of the variance of the original
set. The next best combination, orthogonal to this, would account for less than
4% additional. Hence it appears that the "time" variation is well represented
by a single principal component. Note that the coefficients are all very nearly
1/3 which means that the principal "time" component is essentially just the
average of the 9 individual time parameters.
5. 2. 2 Dose
For the 4 dose parameters the derived coefficient vector was
8 - - . 49
19 - - .51
3 - + . 49
18 - + .51
B-53
LOCKHEED MISSILES & SPACE COMPANY. INC.
-------
LMSC-D406484
This linear combination accounts for 85% of the total variation of the set.
The next best component, orthogonal to this accounts for an additional 14%
and might prove useful. However, mainly for simplicity, we have chosen
to use only the first component, which is equivalent to the average of 3 and
18 minus the average of 8 and 19.
5. 2. 3 MISC I and MISC II
For the 5 remaining parameters the principal component was derived using
initially all 5 and then, again, dropping parameter 16. This was done
because parameter 16 has exhibited some anomalies (Runs 3, 4, and 4A on
aluminum, and Run 42 on Teflon).
It was found that the relative weights among the other four parameters were
similar in both cases and that 16 did not appear to particular contribute to the
desired econimization and it was decided to treat it separately.
The first principal component for the other four parameters was
4 - .56
13 - .48
6 - - .43
9 - .50
This linear combination accounts for 67. 7% of the total variation - the next
component, orthogonal to this accounts for an additional 22% and it was
decided to include this. The second principal components weights are:
4: -.23
13: .54
6: .73
9: . 35
B-54
LOCKHEED MISSILES & SPACE COMPANY. INC.
-------
LMSC-D406484
The first component is simply the sum of 4 + 13 + 9 minus 6.
The second component has no ready interpretation. Henceforth these two
components are called MISC I and MIS II respectively.
Thus, at some potential sacrifice in sensitivity the original 23 parameter set
has been reduced to four: (Because of the time drift, Section 2, PAN Max was
not further considered.) "TIME", "DOSE", MISC I, MISC II. The transformation
used to obtain these parameters is given in Table B5-5. A listing of the four
variables is given in Table B5-6. Time sequential plots of these new para-
meters are presented in Figs. B5-2 thru B5-5.
With only four parameters to keep track of, it is relatively easy to form
scatter diagrams of pairs of parameters, material by material, and this is
presented in the next section, material by material.
5. 3 Results and Interpretation
5. 3. 1 Stainless Steel
Figures B5-6 to B5-11 present scatter plots of the 4 parameters, two at a
time. The experimental conditions for each run are also indicated as follows:
S/V Spectrum Cleaning
High
High
Low
Low
High
High
Low
Low
Full
Full
Full
Full
Cut
Cut.
Cut
Cut
Vacuum
Purge
Vacuum
Purge
Vacuum
Purge
Vacuum
Purge
LOCKHEED MISSILES & SPACE COMPANY. INC.
-------
LMSC-D406484
TIME =
TABLE B5-5
Transformations Used In Obtaining Principal Components
H02™ HCT.50 OoTM FWHM
+ + J +
3(7-35) 3(6.85) 3(8.05) 3(^.83)
+
3(^.18) OJVVG R(3)(.012) HC AVG R(3)(.006)
1 . 1
N02R(3)(-009) NO R (3)(.01)
DOSE = °3DOSE °3DF WOgDOSE NOgDF
2(8.51) 2(1.68) 2(8A3) 2(1.93)
_ _ OoMAXR HC MAX R 0 MAX C HCFC
J. — J , x ,
2(.005T) 2(.005T) 2(033)
MISC II =.
2(.039)
0 MAX R HC MAX R OoMAX C HCFC
J J
.0057)
1.852(.0057) 1-37(033) 2.857 (.039)
Where all variables were originally adjusted for HC Init and % N0_ as in Table B3-1.
B-56
LOCKHEED MISSILES & SPACE COMPANY
-------
LMSC-D406484
Table B5-6
PRINCIPAL COMPONENTS VALUES
ALUMINUM
*UN NO
71
69
85
73
75
81
70
82
8*
74
72
S/V
2
2
2
1
1
2
2
2
2
1
I
REDONE
'UN NO
71
69
85
73
75
81
70
82
84
74
72
T
42
40
40
40
35
53
61
51
55
58
52
SPEC
2
2
2
2
2
1
I
1
1
I
I
ALUMI
IME
.87
.22
.24
.00
.23
.00
.92 -
.63
.60
.24
.60
CLEAN
2
1
I
2
1
2
1
1
1
2
1
NUM
DOSE
-0.55
0.96
-0.04
3.58
5.73
-7.42
10.09
-8.21
-6.92
-4.67
-3.48
HCINIT
3.10
3.04
2.94
3.00
3.06
3.04
2.88
3.07
2.96
3.02
3.00
MISC 1
-5.04
-4.39
-5.34
-9.99
-8.85
-2.06
1.95
-1.43
2.89
-6.3?
-5.67
1
1
I
I
1
1
1
1
1
1
1
NOX
.52
.47
.49
.47
.45
.49
.47
.49
.41
.46
.46
MISC
25.
24.
24.
28.
26.
26.
27.
24.
24.
29.
26.
N02
.25
.21
.13
.14
.19
.14
.17
.17
.14
.16
.15
2
68
56
11
91
46
28
52
47
67
62
50
TN02
16.
14.
8.
9.
13.
9.
11.
11.
8.
12.
10.
4
3
7
5
1
4
6
4
5
3
3
TAOJ
12.
8.
-2.
0.
6.
-2.
4.
4.
0.
6.
0.
HC/NOX
2.00
2.07
1.97
2.0?
2.11
2.04
1.96
2/06
2.10
2.07
2.15
B-57
LOCKHEED MISSILES & SPACE COMPANY. INC.
-------
Table B5-6
PRINCIPAL COMPONENTS VALUES (Cont.)
ORIGINAL ALUMINUM
RUM NO
3
4
4.5
10
11
12
12.5
6
5
9
7
8
OR I
RUN NO
3
4
4.5
10
11
12
12.5
6
5
9
7
8
S/V
2
2
2
1
1
1
I
2
2
I
1
1
SPEC
2
2
2
2
2
2
2
1
1
1
I
I
CLFAN
2
I
I
2
2
I
1
2
1
2
1
1
HCINIT
3.17
3.05
2.90
2.96
3.03
2.98
2.83
2.81
2.98
2.94
2.84
2.92
NOX
1.52
1.54
1.41
1.51
1 .43
1.51
1.67
1.59
1.33
1.49
1.42
1.36
N02
•
•
•
•
•
*
•
•
•
•
•
•
22
29
29
31
22
11
36
31
12
33
22
19
fN02
14
18
20
20
15
7
19
19
9
22
15
14
.5
.8
.6
.5
.4
.3
.2
.5
.0
.1
.5
.0
TAOJ
9.
17.
21.
20.
11.
-8.
18.
28.
0.
35.
16.
12.
HC/NOX
2.01?
1.98
2.06
1.96
2.12
1.97
1.69
?.04
2.24
1.97
2.00
2.15
GINAL ALUMINUM
TIMF
35.
34.
35.
36.
36.
38.
38.
49.
49.
60.
51.
53.
60
19
87
74
61
29
68
87
98
95
73
29
DOSE
0.69
4.48
5.55
8.74
7.59
4.53
1.77
-2.37
0.08
-4.48
-2.11
2.08
MISC I
-4.91
-7.83
-7.12
-7.83
-9.24
-10.04
-6.68
-1.78
2.52
-1.52
-1.60
-3.30
MISC
20.
25.
25.
28.
28.
29.
25.
23.
23.
24'.
24.
26.
2
75
90
69
62
29
21
01
32
6T
14
43
22
ORIGINAL ALUMINUM
B-58
LOCKHEED MISSILES & SPACE COMPANY. INC.
-------
LMSC-D406484
Table B5-6
PRINCIPAL COMPONENTS VALUES (Cont.)
PYREX
RUN NO
22
21
27
33
31
26
25
30
32
S/V
2
2
1
1
1
2
2
1
1
2
2
2
2
2
1
1
1
1
: CL
FAN
2
1
2
2
1
2
1
2
1
HCINIT
3.22
3.20
2.94
3.00
2.84
2.98
2.95
3.23
3.00
NOX
1.68
1.57
1.55
1.52
1.39
1.44
1.54
1.50
1.43
N02
«
«
4
«
4
4
4
4
<
.27
,17
.18
.16
,14
,18
,21
,16
,20
*N02
16.1
10.8
U.6
10.5
10.1
12.5
13.6
10.7
14.0
T40J
12.
0.
5.
0.
0.
7.
10.
2.
11.
HC/NO
1.92
2.04
1.90
1.97
2.04
2.07
1.92
2.15
2.10
PYREX
*UN NO
22
21
27
33
31
26
25
30
32
TIME
44.
43.
31.
33.
41.
79.
93.
62.
52.
38
07
12
26
84
12
19
22
43
005 E
0
-2
11
10
5
-16
-32
-2
-0
.41
.'74
.01
.32
.64
.55
.54
.73
.55
MISC 1
-6.48
-2.52
-11.61
-10.45
-7.84
2.13
12.74
-4.96
-2.81
MTSC
25.
?4.
28.
27.
23.
28.
25.
27.
25.
2
78
42
It
34
83
73
78
53
14
B-59
LOCKHEED MISSILES 8e SPACE COMPANY, INC.
-------
Table B5-6
PRINCIPAL COMPONENTS VALUES (Cont.)
TFFLON
SUN NO S/V
38
34
35
42
43
44
36
39
37
40
41
2
2
2
1
I
I
2
2
2
1
1
TEFLON
SPEC CLEAN
2
2
2
2
2
2
1
1
1
1
I
PUN NO TIME
38
34
35
42
43
44
36
39
37
40
41
51.
44.
45.
52.
47.
50.
76.
80.
70.
79.
88.
05
79
30
39
35
40
34
05
43
27
83
2
1
I
2
1
1
2
2
1
2
I
DOSE
0.
3.
1.
-0.
-1.
-0.
-12.
-15.
-13.
-20.
-29.
25
45
83
91
16
38
92
56
78
17
27
HCINIT
3.00
2.82
2.86
2.97
2.98
3.03
2.99
2.96
2.86
2.96
2.96
MTSC 1
-7.75
-6.80
-6.30
-8.63
-6.54
-6.71
-0.22
-0.01
3.22
1.10
13.25
MOX
1.52
1.38
1.50
1.54
1.52
1.47
1.63
1.45
1.48
1.49
1.47
MISC
27.
26.
26.
33.
31.
29.
30.
29.
26.
31.
30.
N02
•
•
*
•
•
•
•
•
•
•
•
2
95
44
11
49
13
2?
98
39
04
86
56
16
13
15
18
15
15
17
17
15
17
14
SN02
10.5
9.4
14.7
11.7
14.5
13.6
10.4
11.7
10. I
11.4
9.5
TAOJ
0.
-2.
14.
5.
13.
10.
0.
7.
0.
5.
-2.
HC/NOX
1.97
2.04
1.91
1.93
1.96
2.06
1.84
2.04
1.93
1.99
2.01
B-60
LOCKHEED MISSILES & SPACE COMPANY, INC.
-------
LMSC-D406484
Table B5-6
PRINCIPAL COMPONENTS VALUES (Cont.)
STAINLESS STEEL
RUN NO S/V SPEC CLEAN
55
56
51
52
58
62
61
66
57
53
54
63
60
67
STAI
RUN NO
55
56
51
52
58
62
61
66
57
53
54
63
60
67
2 2
2 2
2 2
2 2
2 2
1 2
1 2
I 2
2 1
2 1
2 1
I I
I 1
1 1
NLESS
TIME
25.54
26.88
21.45
21.21
?3.55
31.84
26.51
?8.06
42.24
30.77
34.16
50.19
38.75
38.37
STEEL
2
2
1
1
1
2
1
1
2
1
1
2
1
1
DOSP
4.
3.
6.
6.
5.
9.
8.
10.
-7.
-1.
-2.
-4.
-1.
-0.
87
94
10
45
05
04
57
71
01
96
42
01
83
07
HCTNIT
3.19
3.04
3.09
3.11
2.86
3.02
2.98
3.06
3.11
3.05
3.24
2.88
3.02
3.00
MTSC 1
-6.12
-4.98
-4.00
-3.58
-3.31
-11.02
-9.97
-10.43
2.58
3.34
2.23
-2.20
-1.75
-0.83
NOX
1.48
1.52
1.42
1.43
1.49
1.46
1.50
1.48
1.44
1.49
1.52
1*44
1.55
1.53
MISC
22.
23.
21.
21.
20.
29.
28.
29.
18.
19.
20.
26.
22.
25.
N02
•
•
•
•
•
•
*
•
*
•
•
*
•
•
2
56
11
17
97
97
21
35
04
71
15
19
83
6?
28
15
18
15
15
17
15
16
16
17
16
17
13
18
18
*N02
10.
11.
10.
10.
11.
10.
10.
10.
11.
10.
11.
9.
11.
11.
1
8
6
5
4
3
7
8
8
7
2
0
6
8
TADJ
0.
0.
0.
0.
0.
0.
0.
0.
3.
0.
2.
0.
3.
3.
H:/NOX
2.16
2.00
2.18
2.17
1.92
2.07
1.99
2.07
2.16
2.05
2.13
2.00
1.95
1.96
B-61
LOCKHEED MISSILES & SPACE COMPANY. INC.
-------
10
o<
-10
-20
-30
100
80
60
40
i
9f>
**
C A.
^ •
- FIG. B5-2
DOSE vs RUN SEQUENC
ALUMINUM
HIGH S/V LOW S/V
FIG. B5-3
•— -
•
• _
o 0 c
c
*
: 4
0
C|
c
PYREX
HIGHSAl LOWSA
T 1MB vs RUN SEQUENCE (
—
c
— •
<&
c c c c
v» «**
'c
C!
0'
i
# 1'
^C
I *
1
1 •
\
* e
*,
^
c**^.
c c
;
A
C
<
-.
__^
•-•
*
^
»c
TEFION
1 HIGH SA
s*
C ^-^**
^^^"^
9
C
LOWSA
<
0
c
>c
*
e
•^*
•^ gr -»
^ .
•^
^ c
_
c
^-
•^" ^
c •
c
STAINLESS
HIGH SA
C
c
— ^— «
LOW SA
c
•
c c
•1
1 • 1
• •' 1 — —^ »
1 '
'* • !
1 c 'c
1 l>^p
^ | 1 C
c i !
ALUMINUM RERUNS
LJ|/**LJl * UI^'^LJ
HS^H|LOWSAI HS')^H
C |
^*— i C
1 "~"^~*» •
• ^•sh^
c C
"— ^
um Hi
(ii &
i/a ifi)
• ^ $\ i
fil*
\&m
A C C
5^
5/ C i—
C
—
ALVS RERUNS
HIGH S/Vl LOW SA
AL RERUNS
• A
—
(£\
** w _
A
. A •
u
•e >• sr
w
I
as
ro
-------
10
-10
-20
30
-------
20
Fig. B5-6 STAINLESS
5
o
7;
I
m
m
o
(A
en
F
m
O
m
o
O
2
TJ
>
Z
10
CLEANING
s/v
SPECTRUM
2
-10
Z
n
-20
20
30
40 50
TIME
60
70
03
o
0>
oo
-------
5
O
?;
I
m
m
o
m
0)
n
m
o
0
£
TJ
>
•z.
Z
o
tx)
r
CT\
VJ1
8
6
4
2
0
-2
-4
-6
-8
10
Fig. B5-7 STAINLESS
-12
20
30
40
50
60
70
TIME
o
2
o
05
*.
oo
-------
30
5
n
7;
I
m
m
D
co
en
F
m
(A
U)
•o
>
o
m
n
o
2
TJ
>
Z
Z
n
ON
ON
28
26
U
i 24
22
20
18
Fig. B5-8 STAINLESS
20
30
40 50
TIME
60
70
O
O
00
-------
LMSC-D406484
o>
CO
CM —
U
CM
«o
00
I
o
T
CM
O I
CO
o
CM
3SOO
B-67
LOCKHEED MISSILES & SPACE COMPANY. INC.
-------
LMSC-D406484
o
CM
l/J
.E?
iz
o
o
o
CM
o
CO
o
CO
00
CN
CM
CN
II DSIW
CM
CM
O
CM
00
B-68
LOCKHEED MISSILES & SPACE COMPANY. INC.
-------
LMSC-D406484
to
to
to
o»
CM
OO
OO
o
T
CM
O
CO
00
CM
CM
CM
CM
CM
00
II DSIW
B-69
LOCKHEED MISSILES & SPACE COMPANY. INC.
-------
LMSC-D406484
Thus, for example (+ - +) indicates a run with High S/V, cut spectrum, and
vacuum cleaning.
On each plot, pairs of runs which differ in only one of the three variables
have been connected by arrows which run from the "minus" to the "plus".
If all of the variables has a consistent effect then the arrows will be parallel
for each pair which differs in the same variable. Thus, a very regular
additive model produces a diagram which looks like a three dimensional box.
It is notable that such is the case for all six pictures.
Hence the "orthonalized" analysis presented in Table B5-7 shows very little
interaction for any of the parameters.
5. 3. 2 Teflon
Figure B5-12 thru B5-17 present plots in the same manner as was done for
stainless steel. Here however the pattern is substantially less regular due
almost entirely to the data from Run 41.
We are led to speculate about the validity of this run. It is one of the slowest
runs experienced during the program. Moreover it, together with run 25 on
Pyrex is largely responsible for the initial concern over time trend. Jumping
ahead we note that while run 41 was the final run at cut spectrum - possibly
implying a degradation with lamp and filter age - leading to decreased time;
run 25 was the first run at cut spectrum for pyrex where the apparent trend
is in the opposite direction. Thus there does not seem to be any consistent
trend which might explain the discrepancy. Note also that Run 41 seems to
be discrepant in all 4 parameters.
B-70
LOCKHEED MISSILES & SPACE COMPANY. INC.
-------
s
o
X
X
m
m
o
•
MEAN
A=S/V
B=SPECTRUM
C=CLEANING
AB
AC
BC
ABC
S.S.
33-88
-6.18
-13.72
7.49
.85
.61
-2.97
.32
TIME
Teflon
62.22
-2.08
-25.76
6.01
-.51
.88
-1.26
• 37
Pyrex
52.91
11.47
-25-07
3.14
-4.76
3-07
-7-31
2.4l
Alum
46.97
.94
-14.78
3-11
3.02
-2.09
• 59
1.03
S.S.
1.85
-3.16
10.78
-2.49
-i.o4
- .66
1.45
.22
DOSE
Teflon
-8.00
3-42
16.60
-1.54
-1.13
.10
.28
-1.23
Pyrex
-2.48
-11.48
11-95
.86
-2.16
.56
3.22
-.37
Alum
-1.74
-3.8l
8.11
-1.05
-.39
.37
-.53
-.30
0)
en
m
en
B>
in
o
m
O
O
2
w
MEAN
A=S/V
B=SPECTRUM
C=CLEANING
AB
AC
BC
ABC
S.S.
-3.56
5.23
-8.08
-.97
.80
-.10
-.41
-.45
MISC I
Teflon
-2.67
- .26
-9.44
-2.36
•74
.09
.76
• 31
Pyrex
-3.69
5-94
-6.55
-2.80
-1.01
-.12
-.79
-.26
Alum
-5.66
4.90
-3-85
-1.19
-.42
-.29
.53
• 77
S.S.
MISC II
Teflon
23.91
-6.52
3-24
• 98
-.32
-•72
.02
29.31
-3-4o
• 32
3-12
-1-33
-.22
-.63
1.20
-.61
Pyrex
26.12
.12
-1.11
2.65
-.80
-.49
-.27
-.77
Alum
26.54
-2.66
-.40
2.16
-.02
-.63
-.26
.08
TABLE B5-7 Material Effects for Principal Components
-------
LMSC-D4 06484
z
o
CN
a
o
oo
o
10
o
n
o
CN
o
CN
3SOO
B-T2'
o
CO
-------
LMSC-D406484
8
z
o
CO
2
o>
o
00
o
CO
o
CN
CN
00 -O
CN O CN
I DSIW
•O 00
I I
O CN
B-T3
-------
34
32
30
u
28
26
24
22
Fig. B5-14 TEFLON
20
30
40
50
60
70
80
RUN 41
90
o
os
OO
TIME
-------
RUN 41
w
u
i
10
8
6
4
2
0
-2
-4
-6
-8
-10
-40
«•*
\
Fig. B5-15 TEFLON
\
\
-30
-20
-10
10
DOSE
^~i
8
6
>**
o
£
oo
-------
34,
Fig. B5-16 TEFLON
32
RUN
41
30
u
UJ
5
26
24
22
-40
-30
-20
-10
DOSE
10
20
G
^
o
S
oo
-------
Fig. B5-17 TEFLON
14
12
10
8
6
4
to
I
-q .
-0
U
i
-o
-2
-4
-6
-8
-10
-12
18
20
RUN 41
8
o
O5
OO
22
24
26
MISC II
28
30
32
34
-------
LMSC-D406484
What we have done is visually adjust the data for run 41 so as to bring that diagram
into consistent form as shown. The "adjusted" values of the parameters are
shown on the Figures.
Using the adjusted value, an "orthogonalized" analysis of variance was
calculated and is presented in the same table. Again we note the absence of
interaction in this table.
5. 3. 3 Pyrex
Figures B5-18 thru B5-23 present plots in the same manner as provided for
Teflon and stainless steel. Here the pattern is again substantially less
regular than it was for stainless steel and resembles Teflon in that one
point (Run 25) stands out as different.
As with the Teflon data we are led to speculate about the validity of this run.
It is the slowest run of the entire program, and appears discrepant in three
out of the four parameters. Note that Run 31 also is out-of-line but to a
lesser extent. (Run 31 is a record holder with respect to initial conditions. )
As was done for Teflon the data have been visually adjusted to bring the
diagrams into consistent form.
Using the visually adjusted value for Run 25 (but with no adjustment for Run
31), an orthogonalized analysis was calculated and is shown in the table. As
might be expected, there is, even after adjustment, a sizeable BC interaction
term for both time and dose.
B-78
LOCKHEED MISSILES & SPACE COMPANY. INC.
-------
Fig. B5-18 PYREX
+10
-10
-20
-30
RUN
25
-40
20
30
40
50
60
70
80
TIME
t
90
it
o
*
00
-------
LM8C-D406484
S\
«
\
«
\
•»
\
2
0£
O)
«
\
o
00
o
•o
o
CN
O 00 x>
•O CO
I I
-------
30
Fig. B5-20 PYREX
28
26
bb —
24
22
RUN 31
RUN
25
20
20
30
40
50
60
TIME
70
80
90
G
rf^
o
O5
4^-
oo
-------
bd
fe
u
14
12
10
8
6
4
2
-2
-4
-6
-8
-10
-12
-40
RUN 25
Fig. B5-21 PYREX .
NOTES: LINES CONNECTING CORRESPONDING SPECTRUM PAIRS
ARE OMITTED BUT ARE ALSO "REGULAR"
\
\
\
\
\
*+"
-30
-20
-10
DOSE
10
20
o
O5
00
-------
00
OJ
30
Fig. B5-22 PYREX
28
26
U
to
24
22
RUN 25
RUN 31
20
-40
-30
-20
-10
DOSE
10
20
8
o
Oi
*>•
OO
-------
u
to
14
12
10
8
6
4
2
0
-2
-4
-6
-8
-10
-12
Fig. B5-23 PYREX
20
RUN 31
8
&
o
£
oo
22
24
MISC II
26
28
30
-------
LMSC-D406484
5. 3. 4 Aluminum Reruns
Figures B5-24 thru B5-29 present plots in the same manner as the previous
materials. Here the pattern would be fairly regular were it not for the
substantial variation among the replicate triple - runs 70, 82 and 84 (+--).
It is difficult to justify dropping any one of these 3 runs, since this variation
may in fact be a reflection of experimental error. However, run 70 does
stand out among the three and in forming the diagrams it has been suppressed.
Omitting Run 70, an "Orthogonalized" analysis was calculated and is shown in
the table. Note in the table the general absence of any big interaction terms.
These conclusions are made:
o After judicious pruning of the data, we find nearly additive effects
in all materials (exluding the early aluminum). These are summarized
in Table B5-7.
o The data is effectively summarized in terms of 4 parameters. Of
these - time and dose are highly correlated. The parameters identified
as MISC I and MISC II tend to be associated with maximum values.
See Fig. B5-30.
o In all materials spectrum has a marked effect - cut spectrum runs are
slower, have reduced dose and lower values of maximums (MISC I).
o Teflon and stainless steel are the most different with somewhat
in the middle and aluminum irregularly between and stainless.
o Pyrex is most effected by S/V. High S/V results in increase time,
decreased dose and increased max (MISC I).
o There are only moderate cleaning effects in all materials.
B-85
LOCKHEED MISSILES & SPACE COMPANY. INC.
-------
bd
+10
to
O -10
-20
-30
Fig. B5-24 ALUMINUM
20
30
40 50
TIME
60
70
o
£
oo
-------
ttt
u
to
8
6
4
2
0
-2
-4
-6
-8
-10
-12
20
Fig. B5-25 ALUMINUM
30
RUN
+-- 84
40 50
TIME
60
70
£
4>-
O
«
oo
-------
30
Fig. B5-26 ALUMINUM
28
26
24
RUN 82
84
22
20
30
40
50
60
TIME
70
8
o
05
00
-------
w
I
CD
MD
_ -2
U
i/2
5 -4
-6
-8
-10
-12
Fig. B5-27 ALUMINUM
-40
-30
-20 -10
DOSE
0
10
O
os
-------
w
I
VO
o
30
29
28
27
Fig. B5-28 ALUMINUM
U
25
24
23
22
-40
-30
-20 -10
DOSE
10
8
. o
O5
oo
-------
bd
i
8
6
4
2
Fig. B5-29 ALUMINUM
u
-4
-6
-8
-10
-12
22
24
26
28
MISC
30
32
34
o
£
oo
-------
A STAINLESS
O TEFLON
D PYREX
+ AL (RERUN)
X AL
*
2
0
LU 0
10 ~t-
8-4
-6
-8
-10,
+3
"* X
+
O
—
— 1 °
0 40 50 60 7(
TIME
V
-1
-2
—
u -3
to
i -4
-5
-6
O
A n
X
—
—
30 40 50 60 71
TIME
u
§
ou
29
28
27
26
25
24
93
O
—
-)-
— n
— x
-^
30 40 50 60 70
TIME
u
-1
-2
^ _
U-3
|-4
-5
-6
-7
o
a A
~~ X
—
— 1
1
-10
-2 +2
DOSE
+6
U
to
S
ou
29
28
27
26
25
24
23.
_0
—
—
D +
~~ X
— .
1 A
0 -6 -2 +2 +6
DOSE
JU
29
28
U 27
to
1 26
25
24
9T
O
+ D
X
~ A |
-7-6 -5 -4 -3 -2-1 0 1
MISC I
Fig. B5-30 MEAN VALUES BY MATERIALS
O
rf^
o
oo
-------
LMSC-D406484
5. 4 Univariate Analysis
Several of the analyses reported in the preceeding sections were rather informal
in character. The following sections provide a more formal analysis.
The four new variables were analyzed univariately using the same basic model
given in Section 4. Prior to the analysis of the model, the covariates were
investigated to determine if any were useful (it should be recalled that the data
had been adjusted for HCinit and %NO9), and it was found that NO.,, or HC/NOV
£ .A. .X.
were beneficial to the study of dose, but that if NO,, is included, HN/NO,, is
not necessary. No other dependent variable required any covariate after the
HCinit and %NO? adjustments.
The model proper was then investigated in the same three ways discussed in
Section 4. 1, except that dose parameter was first adjusted for NO.,.. The
results of each analysis are given in Table B5-8. Included in the table are
significant effects for each method of testing. Table B5-9 gives tests of
reduced models in an attempt to obtain a preferred model. The final optimum
model for each dependent variable is given in Table B5-10 with associated
prediction coefficients in Table B5-11. Table B5-12 indicates the significance
of the single degrees of freedom of the significant model terms.
As in Section 4. 1, the results of the univariate analyses are somewhat
camouflaged by the material interactions. Consequently, a "by materials"
study was again performed, with interesting results. All the variables
except MISC II showed a significant spectrum effect, regardless of the
material used. S/V was significant the majority of the time, though a consistent
pattern was not noted. Of most concern was the covariates: NO,., was
'B-93 '
LOCKHEED MISSILES & SPACE COMPANY. INC.
-------
o
n
7s
X
m
m
D
2
U)
F
m
U)
(A
TJ
>
O
m
o
O
z
TJ
>
Z
n
•w
ro ectrum
X
X
X
X
x •
X
X
X
X
X
X
X
X
X
X
X
X
X
.terials
x
x
X
X
X
X
X
X
X
X
X
X
Time In order of model
Time Weighted squares
Time After Mean
Dose In order, after NO
' x
Dose Weighted squares, after
Dose After mean, NO
MISC I In order of model
MISC I Weighted squares
MISC I After mean
MISC II In order of no del
MISC II Weighted squares
MISC II After mean
P
d-
CD
CD
CO
d-
CO
Hj
tf
o
W
vn
i
oo
o
-------
LMSC-D406484
TABLE B5-9
Univariate Tests of Reduced Models
DEP
TIME FULL
TIME RED
TIME W/0 MXC
TIME W/0 C
TBffi W/0 MXC,C
DOSE FULL
DOSE RED
MISC I FULL
MISC I RED
MISC I W/0
MISC II FULL
MISC II RED
MISC II W/0 MXC
DF
30
34
38
35
39
29
33
30
32
36
30
37
4i
D • O • E •
459
489
640
534
707
173
189
90
91
139
47
59
84
.63
• 34
.44
.06
.22
• 30
.14
.819
.430
.10
.882
.934
.231
97
96
95
96
95
96
95
94
94
90
91
89
85
R2
.029
.837
.860
.548
.428
.10956
• 753
.051
.011
.889
.645
.543
• 303
MSB
15
14
16
15
18
5
5
3
2
3
1
1
2
• 321
• 392
.854
.259
.134
•9759
.7316
.0273
.8572
.8638
.5961
.6198
.0644
SS RED
-
29
180
74
247
15
48
12
36
.71
.81
.43
• 59
-
.84
-
.611
.28
.052
• 349
MS
-
7
22
14
27
3
8
1
3
RED
.43
.60
.89
.51
-
• 96
-
• 305
• 05
.722
.304
PREFERRED
F RED MODEL
-
LI
1.48
LI
1.80 *
*
LI *
-
1 *
2.66*
1.08 *
2.07
LOCKHEED MISSILES & SPACE COMPANY. INC.
-------
TABLE B5-10
Optimum Univariate Models on Principal Components
TIME = y* + MAT + SPEC + MAT x S/V + MAT x SPEC + MAT x S/V + SPEC + (.
DOSE = y" + (NOX) + MAT + S/V SPEC + MAT x S/V + MAT x SPEC + MAT
x S/V x SPEC + MAT x CL +£
MISC I =yt + MAT + S/V + SPEC + MAT x S/V + MAT x SPEC + MAT x S/V
x S/V x SPEC + CL + MAT x CL + SPEC x CL + 6
MISC !!=/<+ MAT + S/V + MAT x S/V + MAT x SPEC + S/V x SPEC + CL
+ MAT x CL +£
B-96 i
LOCKHEED MISSILES & SPACE COMPANY. INC.
-------
LMSC-D406484
TABLE B5-11
Coefficients for Univariate Models After Prin Components
CONSTANT
(NOX-1.487)
MATERIAL 1
2
3
4
S/V
SPECTRUM
MAT x S/V 1
.2
3
4
MAR x SP 1
2
3
k
S/V
MAT
x SPEC
x S/V 3
SP
CLEANING
MAT x CL
S/V x CL
SPEC x CL
TIME
DOSE
MISC I
MISC II
1
2
3
4
48.916
_
-^.0573
1.5738
^•3197
6.9811
_
-10.885
-.83580
1.4312
-6.0230'
1.8307
1.1226
-.067122
.18841
-3-9066
_
.17448
.0083518
3-2095
-1.3128
-
_
_
_
-
-2.580
-15.601
1.0637
-1.8314
-2.4776
-3.1876
-1.6585
6.3551
.021104
-.85480
5-3^57
-.38398
-.33^27
.45021
.53569
2.7403
_
-.12435
.51^10
-3.0629
.63217
-
-.43487 .
-.069927
-1.1050
1.0557
-3^336
_
-.055081
-.46526
.50078
1.3192
1.8183
-3-9054
.19419
.68750
-.2637^
-.23285
.020128
.48477
-.87956
-1.2903
_
.12279
-.062527
1.5109
.077586
-1.1976
.13723
-.30636
-16873
-.95245
26.317
_
-.58796
.84906
1.7328
1.0305
-1.4895
-
-.40300
.082604
-.91306
.14814
.34839
-.62572
.26770
-.37077
-.39243
-
-
.
-
.72846
-.04753
.68668
-.071662
.56177
.51965
B-97
LOCKHEED MISSILES & SPACE COMPANY. INC.
-------
IMSC-D40&8I+
TABLE B5-12
Material Contrasts in Optimum Models
TIME DOSE MISC I MISC II
MAT 1 X X • - X
MAT 2 X • • X
MAT 3 XX X
MAT k X X X X
MAT x S/V 1 X X X
MAT x S/V 2 X
MAT x S/V 3 X X X X
MAT x S/V k X
MAT x SP 1 X X
MAT x SP 2 X
MAT x SP 3
MAT x SP k X X XX
MAT x S/V x SP 1
MAT x S/V x SP 2
MAT x S/V x SP 3 X X X
MAT x S/V x SP k
MAT x CL 1 - ' X
MAT x CL 2 - X
MAT x CL 3
MAT x CL 4 - X X X
X Signifies Significance
Contrast Comparison
1 S.S. vs Rest
2 Orig Alum vs Redone Alum
3 Pyrex vs Teflon
k . Orig & Redone Alum vs Pyrex & Teflon
B-98
LOCKHEED MISSILES & SPACE COMPANY. INC.
-------
LMSC-D406484
significant for the time data of stainless steel, and HC/NOX was significant
for the dose data of the redone aluminum. Neither of these covariate results
are consistent with what was expected from previous covariate results. Table
B5-13 summarizes the "by materials" significant terms, Table B5-14 tests
the reduced model for the optimum model, and Table B5-15 presents the "by
material" prediction coefficients for the optimum models.
5. 5 Results of Multivariate Analysis
Repeating the univariate analysis in a multivariate fashion, the covariates
were again tested for significance. As has been prevalent in the previous
multivariate analyses, NO,,, was a significant covariate in the multivariate
analysis. The model given in Section 4 was then analyzed, with NO,., included
.A.
in the model, according to the three approaches presented and Section 4. 1.
As a result of this analysis, the S/V by spectrum and S/V by cleaning inter-
actions were found to be neglible and were consequently dropped from the
model. Table B5-16 gives the results of the significance tests of the effects,
with the optimum model, and Table B5-17 gives the prediction coefficients
associated with the optimum model.
5. 6 Recommended Model
On the basis of the heretofore reported models, the following models are
recommended as best describing the original 23 dependent variables:
Original Alum: Y = M+ /£?HC. .. + & %NO7 + S/V
• ~ \ i init T3, *•
+ spectrum + spectrum x cleaning + £
Pyrex: Y= M+ & HC. .. + H_%NO, • + R NOV
y / rinr^iT rM J r" 5 jf
S/V + spectrum +
B-99
LOCKHEED MISSILES & SPACE COMPANY. INC.
-------
DEP VAR
TIME
S
A
SA
C
SC
AC
COV
DOSE
S
A
SA
C
SC
AC
COV
MISC I
S
A
SA
C
SC
AC
COV
MISC II
S
A
SA
C
SC
AC
COV
TABLE B5-13
BY MATERIALS
ORIG
ALUM
X
X
PYREX
TEFLON
X
X
X
X
X
S.S.
X
X
X
X
NOX
X
X
X
REDONE
ALUM
X
0
X
X
-
_
X
X
_
X
X
X
X
X
X
-
—
HC/NOX
-
X
X
X
X
X
X
X
X
X
-
_
TOTAL
2
5
9
1
O
1
1
k
5
2
1
e
i
i
3
5
2
2
e
i
k
i
e
2
e
e
e
S = S/V
A = SPEC
C = CLEAN
B-100
LOCKHEED MISSILES & SPACE COMPANY. INC.
-------
LMSC-D406484
TABLE B5-14
Tests of Reduced "By Materials" Model
PARAMETER MODEL
TIME
TIME
DOSE
DOSE
MISC I
MISC I
MISC 2
MISC 2
TIME
TIME
DOSE
DOSE
MISC I
MISC I
MISC II
MISC II
TIME
TIME
DOSE
DOSE
MISC I
MISC I
MISC II
MISC II.
TIME
TIME
DOSE
DOSE
MISC I
MISC I
MISC II
MISC II
TIME
TIME
DOSE
DOSE
MAX I
MAX I
MAX 2
MAX 2
Full
Red
Full
Red
Full
Red
Full
Red
Full
Red
Full
Red
Full
Red
Full
Red
Full
Red
Full
Red
Full
Red
Full
Red
Full
Red
Full
Red
Full
Red
Full
Red
Full
Red
Full
"Red
Full.
. Red
Full
Red
ERROR
DF SSE
5
9
5
10
5
10
5
10
2
7
2
6
2
5
2
7
4
9
4
7
4
6
4
8
6.
8
7
8
7
9
7
11
4
9
3
7
4
8
4
9
22.
60.
35.
75.
18.
35*
19-
44.
163.
163.
53.
240.
8.
*7.
1.
7-
45.
.228.
13-
50.
10.
36.
6.
8.
5-
11.
5-
6.
.- 2.
3-
9-
15.
60.
109-
•
2.
12.
21.
6.
17-
566
167
332
362
886
9^3
303
658
89
MSB SS
4
6
7
7
3
3
3
.4
81
2 166
775
76
592
177
586
060
454
88
367
513
051
834
232
446
489
653
571
765
638
442
330
447
777
04
669
118
598
448
166
101
26
40
4
9
1
11
25
3
7
2
.6
l
l
1
l
1
15
12
3
2
1
1
.5113
.6852 37
.0624
.5362 4o
.7772
•59^3 17
.8607
.4658 25
.947
.89 962
.887
.127 186
.296
.435 38
•793
.009 5
.364
.431 183
.3^2
.216 37
• 513
.139 26
.558
.056 2
•91^9
.4566 6
.7958
.8456 l
.3769
.3825
.3328
.4043 6
.194
.116 48
.2231
-.3025 1
.1495
.6810 8
.5414
.9001 10
RED
—
.601
-
.050
-
.057
_
• 355
—
.3
• 98
• 585
-
.474
—
.426
-
.146
-
•783
_
.215
—
.164
_
.194
_
,804
-
.117
—
.263
_
.448
_
.858
_
• 935
OPTIMUM
MS RED F RED MODEL R
—
9
_
8
-
: 3
H
5
—
192
'46
_
12
-
1
—
36
12
_
13
-
«.
3
_
1
_
-
1
—
9
_
2
_
2
.40
.010
.4114
.071
.26
.745
.862
.095
.685
-382
• 392
• 5537
.0546
.1942
.4020
• 5293
.6526
.3621
.2145
.1871
• •
2.1 *
_
2 *
-
1 *
_
P &
„
2.35 *
*
2^.
w
_
3.0 *
-
2 *
—
3-2 *
-
3-7 *
-
5.36 *
_
1 *
—
3-3 *
-
P •£
-
2 .*
_
2 ^
—
1 *
-
2 *
_
1 *
-
2 *
97.50
93-33
80.89
59-21
88.16
77.48
70.76
32.34
95.38
68.25
96.47
84.18
98.05
89.30
93-21
69.74
98.35
91.69
98.85
95.67
97-70
91.57
90.06
86.53
99-41
98.76
98.58
98.27
99-07
98.79
94-51
90.90
92.29
86.17
99-75
99.20
92.31
86.90
81.67
49.15
B-101
LOCKHEED MISSILES & SPACE COMPANY
-------
LMSC-D406484
TABLE B5-15
"By Material" Prediction Coefficients for Univariate Principal Components Data
Const
S/V
Spec
| S/V x Spec
fj Clean
S/V x CL
Spec x CL
(NOX - 1.482)
Const
S/V
H Spec
§ S/V x Spec
C Clean
S/V x CL
Spec x CL
(HC/NOX-2.05)
Const
S/V
H Spec
o S/V x Spec
H Clean
S S/V x CL
Spec x CL
Const
H S/V
u Spec
H S/V x Spec
S Clean
S/V x CL
Spec x CL
Orig
Alum
44.557
-1.8053
-8.2458
-
_
_
-
-
1.7026
_.
3.0618
-
-
_
_
-
-4.4090
_
-3.2552
_
_
-
25.206
_
-1.3525
-
-
-
-
Pyrex
55-235-
_
-16.502
-
-
_
-
-
-4.8816
-7.9752
8.2119
-
-
_
_
-
-2.3872
3.8556
. -4.1634
_
-2.2830
_
-
26.146
_
-
-
1.3525
-
-
Teflon
63.766
_
-15 .218
-
-
-
-
-
. -9.444
3-3249
9.9576
-1.9924
-
_
-
-
-1.4645
_
-5-9243
1.1883
. -2.2897
_
1.4864
29.678
-1.86i6
-
-
1.4267
-
-
s.s.
33.937
-2.8295
-6.9444
-
4.0072
-
-2.o46i
17.003
1.8344
-1.4769
5.373^
-.56349
-1.2735
_
.69670
-
-3-6053
2.6i43
-3-9788
.46127
-.56056
-
-
23-733
-3.1543
1.6028
-
-
-
-
Rerun
Alum
47.606
-
-7.8917
-
-
-
-
-
-1.985
-1.5782
4.5361
-
-
-
-
18.618
-4.9768
2.7306
-2.2917
-
-
-
-
26.601
-1.2740
-
-
-
-
-
B-102
LOCKHEED MISSILES & SPACE COMPANY
-------
TABLE B5-16
Multivariate Test of Effects
Material
S/V
Spectrum
Mat x .
Mat x Spec
S/V x Spec
Mat x S/V x Spec
Cleaning
Mat x Clean
S/V x Clean
Spec x Clean
NO : F = 5-
x
X
CQ
-------
IMSC-D406484
TABLE B5-17
Coefficients for Multivariate Model After Principal Components
Model Term Time Dose Max I Max 2
CONSTANT
(NO -1.487)
MATX1
MAT 2
MAT 3
MAT 4
S/V
SPECTRUM
MAT 1 x S/V
MAT 2 x S/V
MAT 3 x S/V
MAT k x S/V
MAT 1 x SPEC
MAT 2 x SPEC
MAT 3 x SPEC
MAT k x SPEC
MAT 1 x S/V x SPEC
MAT 2 x S/V x SPEC
MAT 3 x S/V x SPEC
MAT 4 x S/V x SPEC
CLEANING
MAT 1 x CLEAN
MAT 2 x CLEAN
MAT 3 x CLEAN
.MAT k x CLEAN
SPEC x CLEAN
49.218
3.8289
-3.8695
1.5168
4.2159
6.9024
.3619
-10.917
-8-534
1.5761
-5-9708
1.7102
1.0667
-.0308
.3168
3-7014
.1485
-.1155
3.4650
-1.3544
.9283
.6633
.0201
1.5937
-.5485
-.4320
-2.514
-18.589
1.0743
-1.8243
-2 .4491
-3-1933
-1.6058
6.4115
-.0015
-.7846
5.2283
-.3885
-.3556
.4053
.5818
2.7594
-.1322
.5000
-3.1403
.6550
.5660
-.4801
-.0303
-1.1534
.1.1069
.0411
-3.437
2-5979
-.0504
-.4564
.5229
1.2806
1.7952
-3.94l6
.2054
.6714
-2.5770
-.2518
.0353
.6208
-.8339
-1.2875
.1214
-.0923
1-5754
.0638
-1.2382
.1543
-.2994
.1750
-.9697
• 5143
26.310
-3.6848
-.6104
.8848
1.7100
1.0934
-1.4738
• 3075
-.4316
.1449
-1.0087
.1651
• 3107
-.7602
.1426
-.3104
-.0071
.2866
-.1101
.0339
.8006
-.0803
.6460
-.0244
.6234
-.0015
B-104
LOCKHEED MISSILES & SPACE COMPANY
-------
LMSC-D406484
Teflon: Y = JU + f^HCinit + $3.%N°2 + S/V
+ spectrum + S/V x spectrum 4"
Stainless Steel: Y = A.4+ f HCinit + ^°NO2 +
S/V + spectrum + cleaning + £
Redone Alum: Y=, U+ Q HCinit + fc/° T°2 + $3HC/N°X
+ S/V + spectrum +
This recommendation is made after consideration of the univariate and multi-
variate analyses, both before and after the principal components have been
used. It is felt that these models are the most implicated models
as being descriptive of the data.
The manner in which this analysis proceeded was as follows:
o Covariance corrections were calculated for the 23 parameters
separately.
o A multivariate analysis of covariance was used to check that the
covariates selected were appropriate.
o Using the covariate corrected data, principal components for TIME,
DOSE, MISC I and MISC II were developed.
o These new parameters were reanalyzed using both univariate and
multivariate analysis of covariance.
o Somewhat less formal techniques revealed that certain
especially #25 and #41 were possible outliers. Suppression of
these points led to a substantially simplier model.
The outliers had an opportunity to influence, perhaps strongly influence, the
covariance corrections. Thus to be completely consistent they should be
suppressed arid the entire analysis redone. This has not been done both for
B-105
LOCKHEED MISSILES & SPACE COMPANY. INC.
-------
LMSC-D406484
reasons of time and budget and also because it is felt that further effort
should be devoted towards deriving better descriptors of the original data.
It is clear that for this data the parameters were excessively redundant.
It is possible that some features of the original graphs have been over-
looked and this merits investigation.
B-106
LOCKHEED MISSILES & SPACE COMPANY. INC.
------- |