Ecological Research Series
EFFECTS OF GASEOUS
POLLUTANTS ON MATERIALS
A Chamber Study
PRO&
Environmental Sciences Research Laboratory
Office of Research and Development
U.S. Environmental Protection Agency
Research Triangle Park, North Carolina 27711
-------
RESEARCH REPORTING SERIES
Research reports of the Office of Research and Development, U.S. Environmental
Protection Agency, have been grouped into five series. These five broad
categories were established to facilitate further development and application of
environmental technology. Elimination of traditional grouping was consciously
planned to foster technology transfer and a maximum interface in related fields.
The five series are:
1. Environmental Health Effects Research
2. Environmental Protection Technology
3. Ecological Research
4. Environmental Monitoring
5. Socioeconomic Environmental Studies
This report has been assigned to the ECOLOGICAL RESEARCH series. This series
describes research on the effects of pollution on humans, plant and animal
species, and materials. Problems are assessed for their long- and short-term
influences. Investigations include formation, transport, and pathway studies to
determine the fate of pollutants and their effects. This work provides the technical
basis for setting standards to minimize undesirable changes in living organisms
in the aquatic, terrestrial, and atmospheric environments.
This document is available to the public through the National Technical Informa-
tion Service, Springfield, Virginia 22161.
-------
EFFECTS OF GASEOUS POLLUTANTS
ON MATERIALS—A CHAMBER STUDY
by
F. H. Haynie
J. W. Spence
J. B. Upham
Environmental Sciences Research Laboratory
Research Triangle Park, North Carolina, 27711
U.S. ENVIRONMENTAL PROTECTION AGENCY
OFFICE OF RESEARCH AND DEVELOPMENT
ENVIRONMENTAL SCIENCES RESEARCH LABORATORY
RESEARCH TRIANGLE PARK, NORTH CAROLINA, 27711
-------
DISCLAIMER
This report has been reviewed by the Environmental Sciences Research Lab-
oratory, U.S. Environmental Protection Agency, and approved for publication.
Mention of trade names or commercial products does not constitute.endorsement
or recommendation for use.
11
-------
ABSTRACT
This document describes a comprehensive laboratory study using specially
designed controlled environment exposure chambers to assess the effects of
gaseous air pollutants (sulfur dioxide, nitrogen dioxide, and ozone) on a
variety of materials. Materials included weathering steel, galvanized steel,
aluminum alloy, paints, drapery fabrics, white sidewall tire rubber, vinyl
house siding, and marble. The exposure experiment was statistically designed
using a two-level factorial arrangement to identify the environmental factors
or combination of factors, or both, that cause materials damage. Over 200
different direct and synergistic effects were examined. The study revealed
that only 22 of the possible effects were statistically significant at better
than a 95 percent confidence level. Sulfur dioxide, relative humidity, and
the-interaction between them, were the main factors causing effects. A number
of empirical functions were developed that relate materials effects to various
factors causing the effects. An exceptionally good relationship was obtained
for the corrosion of weathering steel.
The lack of statistical significance that was found for the large majority
of effects that were studied is equally as important as the significant effects.
As a result a large number of material-pollutant combinations may be excluded
from'further detailed study.
111
-------
CONTENTS
Page
ABSTRACT iii
LIST OF FIGURES vii
LIST OF TABLES ix
I INTRODUCTION 1
II SUMMARY 2
III CONCLUSIONS 4
IV RECOMMENDATIONS 5
V EXPOSURE SYSTEM 6
Design Features
Evaluation of the System
Chamber Lighting
Chamber Pollutant Distribution
Control Capability of Environmental Factors
VI STUDY DESIGN 12
Selection of Materials
Statistical Design of Exposure Experiment
Treatment of Data
VII MATERIALS TESTING TECHNIQUES AND EVALUATION 17
Testing Techniques
Evaluation
VIII EFFECTS ON WEATHERING STEEL 21
Methodology
Results and Discussion
IX EFFECTS ON GALVANIZED STEEL 32
Methodology
Results and Discussion
X EFFECTS ON ALUMINUM ALLOY 38
Methodology
Results and Discussion
v
-------
CONTENTS
Page
XI EFFECTS ON PAINTS 47
Methodology
Results and Discussion
XII EFFECTS ON DYED FABRICS 59
Methodology
Results and Discussion
XIII EFFECTS ON ELASTOMERS AND PLASTICS 71
Methodology
Results and Discussion
XIV EFFECTS ON MARBLE AND CEMENT 79
Methodology
Results and Discussion
XV REFERENCES 84
VI
-------
LIST OF FIGURES
Figure Page
1. Exposure System Flow Diagram 7
2. Temperature-Time Profiles for Exposed Metal Panels during
Chamber Dew/Light Cycles 25
3. Comparison of Predicted and Measured Thickness-Loss Values
for Weathering Steel Exposed to Laboratory Controlled Polluted
Air and Clean Air Conditions 29
4. Comparison of Predicted and Actual Thickness-Loss Values for
Galvanized Steel Exposed to Laboratory Controlled Polluted
Air and Clean Air Conditions 35
5. Scanning Electron Photomicrograph of an Exposed Galvanized
Steel Panel Showing Uniformly Dispersed Crystalline Material
Over the Zinc Surface 37
6. Method of Stressing C-Ring Specimens 39
7. Wiring of Specimens to Record Times-of-Failure 40
8. Stress Induced Intergranular Cracks in Aluminum Alloy 7005-T35
after 1000 Hours Exposure to Air Containing S02 41
9. Fracture Face Cross Section of Aluminum Alloy Specimen
Ruptured by Bending after Intergranular Cracks Developed
During 1000 Hours Exposure to Air Containing S02 42
10. Stress Corrosion Crack after 2000 Hours Exposure to Clean
Air 46
*
11. Comparison of Predicted and Measured Fading Values for the
Plum Colored Fabric Exposed to Laboratory Controlled
Pol luted Air and Clean Air Conditions 68
12. Effect of Nitrogen Dioxide Concentration on the Fading of
the Plum Colored Fabric 70
13. Macrographs of Cracks Developed by White Sidewall Rubber
Tire Specimens when Exposed Under High Stress to Designated
Controlled Polluted Air Environmental Conditions for
1000 Hours 74
VI1
-------
LIST OF FIGURES
Figure Page
14. SEM Photomicrograph (500K) of Marble Specimen Exposed for
1000 Hours to High Levels of Pollutants and High Relative
Humidity 81
15. Sulfur X-Ray Microprobe Scan (500K) of Marble Specimen
Shown in Figure 14 , 81
Vlll
-------
LIST OF TABLES
Table Page
1. Statistically Significant Factors that Damaged Materials
Exposed to Controlled Polluted Air Environmental Conditions 3
2. Analysis of Variance of Xenon Lamp Energy Data.
3. Analysis of Variance of Pollutant (N02) Distribution within
the Chambers 9
4. Control Capability of Environmental Factors within the Chambers 10
5. Ranking of Materials According to Estimated Damage by Gaseous
Pollutants 12
6. Materials Selected for the Chamber Exposure Study 13
7. Environmental Factors and Levels Used in the Chamber Exposure
Experiment ! 14
8. Two-Level Factorial Arrangement 15
9. Materials Testing Techniques 17
10. Rate of Loss of Film Thickness During the Dew/Light Cycle
Exposure Condition •> 18
11. Rate of Loss of Film Thickness During the Constant Temperature
and Humidity Exposure Condition 18
12. Analysis of Variance for Rayon Tire Cord Breaking Strength 19
13. Elemental Analysis of Weathering Steel 21
14. Corrosion Rates of Weathering Steel Exposed to Designated
Controlled Polluted Air Environmental Conditions 22
15. Corrosion Rates of Weathering Steel Exposed to Designated
Controlled Clean Air Environmental Conditions 23
16. Analysis of Variance of Weathering Steel Corrosion Data
for Polluted Air Exposure Conditions 24
17. Analysis of Variance of Weathering Steel Corrosion Data
for Clean Air Exposure Conditions 24
ix
-------
LIST OF TABLES
Table Page
18. Time-of-Panel-Wetness per Dew/Light Cycle and Geometric
Mean Panel Temperature when Wet as a Function 'of Input
Temperature and Relative Humidity 26
19. Factors Affecting Atmospheric Corrosion of Weathering Steel
Exposed at Urban Sites • 30
20. Corrosion Predictability as Measured by Coefficients of
Variation, R2, between Predicted and Field Values 31
2l. Corrosion Rates of Galvanized Steel Exposed to Designated
Controlled Polluted Air Environmental Conditions 33
22. Analysis of Variance of Galvanized Steel Corrosion Data
for Polluted Air Exposure Conditions 33
23. Elemental Analysis of Aluminum Alloy 7005-T53 38
24. Bending Strength of Aluminum Alloy Stress Corrosion Specimens
after Exposure to Designated Controlled Polluted Air
Environmental Conditions for 1000 Hours 43
25. Bending Strength of Aluminum Alloy Stress Corrosion Specimens
after Exposure to Designated Controlled Clean Air Environmental
Conditions for 1000 Hours 44
*
26. Nominal Paint Film Thickness on Exposed Panels 47
27. Factors for Converting Loss-in-Weight to Loss-in-Film-Thickness 49
28. Erosion Rates of Oil Base House Paint Exposed to Designated
Controlled Polluted Air Environmental Conditions 50
29. Erosion Rates of Oil Base House Paint Exposed to Designated
Controlled Clean Air Environmental Conditions 50
30. Analysis of Variance of Oil Base House Paint Erosion Rates
for Polluted Air Exposure Conditions 51
31. Erosion Rates of Vinyl Coil Coating Exposed to Designated
Controlled Polluted Air Environmental Conditions 54
-------
LIST OF TABLES
Table Page
32. Erosion Rates of Vinyl Coil Coating Exposed to Designated
Controlled Clean Air Environmental Conditions 54
33. Analysis of Variance of Vinyl Coil Coating Erosion Rates
for Polluted Air Exposure Conditions 55
34. Erosion Rates of Acrylic Coil Coating Exposed to Designated
Controlled Polluted Air Environmental Conditions 56
35. Erosion Rates of Acrylic Coil Coating Exposed to Designated
Controlled Clean Air Environmental Conditions. 57
36. Analysis of Variance pf Acrylic Coil Coating Erosion Rates
for Polluted Exposure Conditions 58
37. Description of Drapery Fabrics 59
38. Initial Fading Rates of Drapery Fabrics Exposed to Designated
Controlled Polluted Air Environmental Conditions 61
39. Initial Fading Rates of Drapery Fabrics Exposed to Designated
Controlled Clean Air Environmental Conditions 62
40. Analysis of Variance of Royal Blue Fabric Fading Rates for
Pol luted Air Exposure Conditions 63
41. Analysis of Variance of Red Fabric Fading Rates for Polluted
Air Exposure Conditions , 64
42. Analysis of Variance of Plum Fabric Fading Rates for Polluted
Air Exposure Conditions 65
43. Analysis of Variance of Fading Rates for Clean Air Exposure
Conditions 66
44. Cord Breaking Strength of White Sidewall Rubber Tire Specimens
Exposed under Stress to Designated Controlled Polluted Air
Environmental Conditions for 1000 Hours 73
45. Analysis of Variance of Tire Cord Breaking Strength for
Statistically Significant Factors 74
-------
LIST OF TABLES
Table Page
46. Cracking Rates of White Sidewall Rubber Tire Specimens
Exposure under Stress to Designated Controlled Polluted
Air Environmental Conditions ; 76
47. Analysis of Variance of Rubber Cracking Rates for
Statistically Significant Factors.... 76
48. Erosion Rates of Marble Exposed to Designated Controlled
Polluted Air Environmental Conditions 82
49. Analysis of Variance of Marble Erosion Rates for Polluted
Air Exposure Conditions 82
XI1
-------
SECTION I
INTRODUCTION
The Clean Air Amendments of 19701 call for research on the effects of air
pollutants on materials. Such information is needed to serve as input for
cost-benefit studies and as criteria for developing secondary air quality
standards. As the first step in meeting this mandate, EPA investigators con-
ducted surveys of the literature and of industrial organizations to identify
and assess known effects of air pollutants on various materials. The result
was a compilation of a number of documented materials-effects problems. The
type of information documented, however, was too general and lacked such neces-
sary details as dose-response relationships. Comprehensive laboratory research
and supporting field studies were therefore deemed necessary.
This report describes the results of an initial laboratory study to
statistically identify the direct and synergistic effects of gaseous pollutants
(sulfur dioxide, ozone, and nitrogen dioxide), temperature, and humidity on
several classes of materials. The laboratory study was conducted in specially
designed controlled environment exposure chambers. Further laboratory and
field exposure studies, based on the results of this initial study were to be
conducted in order to develop damage predictive equations from dose-response
data. These plans, as well as all materials research, however, have been
terminated because of budgetary limitations and low priority.
-------
SECTION II
SUMMARY
Fourteen different economically important materials were exposed to 16
different combinations of S0£, 03, N02, and relative humidity levels in laboratory
controlled environment chambers. The environmental factors and twelve of the
materials are presented as a matrix in Table 1. Specimens of latex house
paint and cement were also exposed but the damage assessment technique pro-
duced unuseable data. The aluminum alloy and white sidewall tire rubber were
exposed at two different stress levels to study the direct and synergistic
effects of that factor on those two materials.
The statistically designed two-level factorial experiment made it possible
to study 242 different direct and synergistic environmental effects. Poor data
for the latex house paint and the cement reduced the number of possible effects
to 212. The study revealed that only 22 of the possible effects were statis-
tically significant at better than a 95 percent probability level. These effects
are indicated in Table 1.
It is equally important to note the effects that were not statistically
significant, because these represent possible areas of study that no longer
need to be considered. For example, no significant effects were observed on
the acrylic coil coating and the vinyl siding, and only relative humidity
affected the blue and red fabrics. Thus, negative results would be expected
from any further studies of these materials and pollutants.
When the magnitude of damage and the economics of material selection and
replacement are considered, the vinyl coil coatings, the plum fabric, and the
white sidewall tire rubber are not appreciably affected by these pollutants.
The remaining materials are affected mainly by SC>2, relative humidity, and the
interaction between these two factors.
-------
Table 1. STATISTICALLY SIGNIFICANT FACTORS THAT DAMAGED MATERIALS
EXPOSED TO CONTROLLED POLLUTED AIR ENVIRONMENTAL CONDITIONS
SO, x RH x 03 x N02
J A Significant at the 99 and 95 percent probability levels, respectively.
RH,'' Relative humidity
x means interacting with
NOTE: Both the aluminum stress corrosion specimens and the rubber were
stressed at two different levels, thus the effects of stress and
the interactions of stress with the 15 above factors were also
possible, adding 32 additional possible effects.
-------
SECTION III
CONCLUSIONS
Based on the results of a statistically designed controlled environment
chamber study it is concluded that:
1. Sulfur dioxide at ambient levels contributes significantly to the
corrosion of weathering steel and galvanized steel.
2. Sulfur dioxide at ambient levels contributes significantly to the
erosion of oil base house paints containing calcium carbonates and
marble.
3. Sulfur dioxide at ambient levels contributes significantly to the
stress-corrosion cracking of 7005 T-6 aluminum alloy.
4. Vinyl and acrylic coil coatings, two drapery fabrics, white sidewall
tire rubber and vinyl house siding were not significantly affected,
either Statistically or economically, by S02, N02, or 03 at ambient
levels.
5. The damage assessment techniques used for the latex house paint and
the cement produced data too poor to analyze.
4
-------
SECTION IV
RECOMMENDATIONS
Both laboratory and field studies should be initiated to determine dose-
response relationships for the effects of sulfur dioxide on the following
materials:
1. Weathering steel
2. Galvanized steel
3. Oil base house paint
4. Latex house paint
5. 7000 series aluminum alloy under stress
6. Marble
7. Cement
The laboratory study should be statistically designed to include five
levels each for S02, temperature, relative humidity, and time. The S02 levels
should range around existing annual average ambient air quality standards.
Field studies should be long enough to show non linear accumulative
effects of S02.
-------
SECTION V
EXPOSURE SYSTEM
DESIGN FEATURES
Controlled environment chambers provide the best means for generating
materials effects data. For this study a sophisticated exposure system
consisting of five chambers was designed and built.2 The chambers operated
at temperatures, relative humidities, and pollutant levels normally found to
exist in ambient environments. To accelerate the environmental effects, how-
ever, a programmed dew/light cycle became a unique feature of each chamber.
The dew/light cycle simulated day/night conditions, but could be repeated
many times during a 24-hour period.
Figure 1 shows the basic flow diagram of the exposure system. Ambient
air, after being filtered to remove particulate (>5y) and gaseous pollutants,
was cooled and dehumidified. The conditioned, clean air then flowed into
five separate ducts (one for each chamber), each containing heater and steam
humidification units, which reconditioned the air to a desired temperature and
relative humidity. From each duct the air entered a mixing box that housed
temperature and humidity control sensors and three gaseous pollutant (S02, N02,
03) injection ports. The reconditioned, polluted air next flowed into the
base of the chambers, across a plenum to facilitate mixing, up over the test
specimens, out of the chambers into a decontamination unit, and finally to
the outside.
To provide for a dew/light cycle, each exposure chamber featured a xenon
arc lamp (6000 W) to simulate sunlight and chill racks upon which material
test specimens were mounted. With the lamp off, collant circulated through
the racks, thus cooling the specimens and resulting in the formation of dew.
When the collant stopped circulating, the lamp came on and the dew evaporated
from the specimens. To insure increased thermal conductivity between test
specimens and racks, silicone paste was applied to the backsides of most
specimens. The paste also served to hold the specimens in place and protect
the backsides from exposure.
Automatic instrumentation independently controlled the concentrations
of the gaseous pollutants within each chamber. Air temperature and relative
humidity were controlled prior to flowing into the chambers. During the
dew/light cycle, however, both the air temperature and relative humidity
fluctuated in the chambers.
EVALUATION OF THE SYSTEM
Results of a statistically designed controlled-environment experiment
will only be as good as the controls. Thus, variability from the desired
-------
AIR INLET
t
ROUGHING
FILTER
PARTI CULATE
FILTER
CHARCOAL
FILTER
DEHUMIDIFYING/
COOLING SYSTEMS
CENTRIFUGAL
BLOWER
HEAT AND STEAM
(HUMIDITY)
MANIFOLD INTO 4
OTHER
ENVIRONMENTAL
UNITS
MIXING BOX
HEAT/HUMIDITY CONTROLS
GASES: S02. N02, 03
INJECTION PORT
CHAMBER
MAN I FOLD FROM 4
'OTHER CHAMBERS
CHARCOAL
FILTER
PURAFIL
FILTER
EXHAUST TO
ATMOSPHERE
Figure 1. Environmental system flow diagram.
-------
levels of pollutants, temperature, relative humidity, and light energy should
be minimized. Statistical techniques were used to determine which parts of
the system needed adjustment.
Chamber Lighting
The energy distribution from the xenon lamp within each chamber was
measured and recorded as millivolts by a Talley Industries SOL-A-Meter. The
distribution of energy on the specimen racks within the five chambers was
balanced by (1) varying the lamp wattage, (2) placing reflectors in the light
cap, (3) varying the angle of the specimen rack, and (4) reducing the reflec-
tivity of the walls with a light spray of flat black paint.
An analysis of variance (Table 2) was carried out on the energy data to
determine chamber and position effects. The calculated F statistic (Table 1)
revealed that chamber differences were statistically insignificant. The
position effect, however, was significant at the 5 percent probability level.
Best estimate of the overall mean energy (T) was 61.51 millivolts with a
standard deviation (2S) of 3.38 for 89 degrees of freedom. The coefficient
of variation was 3 percent. With these tolerance limits it was 95 percent
certain that less than a 6.8 percent variation from the mean energy existed
for 95 percent of the measurements. This relatively small amount of vari-
ability did not warrant the stratification of position as a variable.
Placing the material test specimens randomly within each chamber minimized
any bias that could have been caused by the position effect.
Table 2. ANALYSIS OF VARIANCE OF XENON LAMP ENERGY DATA
Source
Position
Chamber ,
Residual
Total
Sum of
squares
55.5960
5.7005
89.0596
150.3560
Degrees of
freedom
17
4
68
79
Mean
squares
3.2704
1.4251
1.3097
1.6894
calc
2.05
1.09
Ftablea
1.76
2.50
5 percent probability level.
Residual was confounded with a possible chamber x position interaction
effect because the experiment was not replicated. It was taken as the
error term in calculating the F values.
-------
Chamber Pollutant Distribution
It is essential that the movement of polluted air be uniformly distributed
across the test specimens. A plenum (sheet of stainless steel with 0.74-cm
holes space 2.5 cm apart) was installed about 12.7 cm above the base of each
chamber. The plenum created a pressure drop in the movement of air across
it, thereby facilitating the mixing of pollutants within the chambers. The
air flow from each chamber was measured in the exhaust ducts with a pitot
tube. The air flow to each chamber was then balanced by means of a vane
installed in the air supply duct.
To measure the distribution pattern of polluted air within the chambers,
specimens of blue-colored test ribbon were placed at various locations within
the chambers about 15.24 cm above the plenum and parallel to the air flow.
The test ribbon was developed by the American Association of Textile Chemists
and Colorists and is sensitive to N02- All five chambers were operated at
constant conditions:
Temperature
Relative humidity
Nitrogen dioxide
Air flow
35°C (95°F)
5 percent
940 yg/m3
2.7 nr/min (2.5 air changes per minute)
After 48 hours exposure, the color of the ribbon specimens was measured
photoelectrically with a Hunter Model D25A Color Difference Meter. The
magnitude of the color change of each ribbon specimen was recorded as AE
values. An analysis of variance (Table 3) was conducted to determine the
significance of chamber and position effects.
Table 3. ANALYSIS OF VARIANCE OF POLLUTANT (N02)
DISTRIBUTION WITHIN THE CHAMBERS
Source
Position
Chamber ,
Residual
Total
Sum of
squares
0.02344
0.00424
0.02441
0.05209
Degrees of
freedom
8
4
32
44
Mean
squares
0.00293
0.00106
0.00076
0.00118
calc
3.84
1.40
Ftablea
2.25
2.67
5 percent probability level.
Residual was confounded with a possible chamber x position interaction
effect because the experiment was not replicated. It was taken as the
error term in calculating F values.
-------
The calculated F statistic (Table 3) for the pollutant distribution
within the chambers indicated that the chamber effect was statistically
insignificant; however, position effects existed at the 5 percent probability
level. The best estimate of the overall mean (x) for the AE values was
3.50 with a standard deviation (2S) of 0.069 for 8 degrees of freedom. The
coefficient of variation was 1 percent. With these tolerance limits it was
95 percent certain that less than a 2.25 percent variation from the mean
color change could be expected for 95 percent of the measurements. Again,
randomly placing the test specimens within the chambers minimized bias that
this variable could have caused.
Control Capability of Environmental Factors
Table 4 shows the control capability of the environmental factors—tem-
perature, humidity, and pollutants--based on 24 hours continuous exposure
and recording. The variability (2S) for the control capability of the
Table 4. CONTROL CAPABILITY OF
ENVIRONMENTAL FACTORS WITHIN THE CHAMBERS
Control Capability
Environmental factors Control point x 2S
Temperature, C°
Humidity, %
Ozone, ug/m3
Sulfur dioxide, yg/m3
Nitrogen dioxide, pg/m3
35 (95°F)
90
50
980
196
1310
262
940
34.8
88.8
50.4
991.8
199.9
1372.9
275.1
930.6
+2
+2.2
+1.1
+31.4
+ 11.8
+131.0
+23.6
+26.3
environmental factors appeared to be proportional to some function of level
Cx). Such a relationship is not uncommon and is frequently encountered in
controlled experimentation. It posed no problems in computations using
exposure data.
The control of temperature and relative humidity was within the mixing
box prior to the air entering the environmental chambers. During the dew/light
10
-------
cycle, the temperature and relative humidity were not controlled but allowed
to fluctuate in the chambers. The three gaseous pollutants were injected
into the mixing box but were controlled within the chambers. Air samples were
continuously taken from each of the chambers by means of a manifold system
connected to a monitoring analyzer. The monitoring analyzers were specific
for measuring concentrations of each of the three pollutant gases. The levels
of the pollutants were constant during the dew/light cycle.
11
-------
SECTION VI
STUDY DESIGN
The overall study was a first-cut attempt to determine under simulated
conditions the effects on materials of three gaseous pollutants for which
national ambient air quality standards have been established. To accomplish
this goal, a statistical experimental design was chosen to provide data for
assessing both direct and synergistic effects.3 Parameters selected and
incorporated into the experimental design included materials for exposure,
environmental factors (temperature, humidity, pollutants), and levels of
control.
SELECTION OF MATERIALS
In prior years, EPA sponsored state-of-the-art surveys to gather technical
and economic information concerning the effects of air pollution on material
products. A systems approach study,** prepared by Midwest Research Institute
(MRI), ranked materials according to potential economic damage by gaseous air
pollution. Table 5 summarizes the results of the ranking by generic classes
of materials.
Table 5. RANKING OF MATERIALS ACCORDING TO
ESTIMATED DAMAGE BY GASEOUS POLLUTANTS
Number of
Dollar loss, references with
Materials percent of total useful information
Metals
Paints
Textiles
Elastomers
Plastics
All others
39
32
9
5
3
12
54
7
14
12
9
4
The potential economic loss for pollutant-damaged metals ranked higher than
other materials; published information was also highest. Damage to paints
ranked second in economic importance. Equally significant, the MRI study
revealed that only a small amount of published information was available on
the effects of air pollutants on paints and other non-metallic materials.
12
-------
Using the MRI study as a guide, one or more materials from each generic
class were selected for the exposure study. Table 6 lists these materials;
they all represent large sales volume and therefore are commercially important.
Table 6. MATERIALS SELECTED FOR
THE CHAMBER EXPOSURE STUDY
Metals Weathering steel, galvanized steel, aluminum alloy
Paints Oil base house paint, acrylic latex house paint,
vinyl coil coating, acrylic coil coating
Textiles Dyed drapery fabrics
Elastomers White sidewall rubber from a radial ply tire
Plastics Vinyl house siding
Stone White Cherokee marble
Cement Portland
STATISTICAL DESIGN OF EXPOSURE EXPERIMENT
A statistically designed exposure experiment must include environmental
factors that are most likely to affect the service life of materials. Environ-
mental factors known to affect materials during exposure are sunlight (ultra-
violet radiation), relative humidity, temperature, dew, and pollutants. Other
environmental factors such as snow and frost occur infrequently with seasonal
changes and are difficult to simulate in controlled environmental chambers.
Of the gaseous pollutants for which national ambient air quality standards
have been set, sulfur dioxide (S02), nitrogen dioxide (N02), and ozone (03)
are known to affect different materials during their expected service life.
As many of these environmental factors as possible, therefore, should be
incorporated into a statistically designed experiment.
Although numerous statistical plans for conducting experiments and
collecting data are available, a two-level factorial arrangement was selected
because it is an excellent method for identifying the environmental factors
or combination of factors, or both, that produce significant effects (materials
damage). Table 7 presents the environmental factors and exposure levels
selected. The low levels for the three gaseous pollutants represent the
national primary ambient air quality standards, whereas the high levels
represent concentrations that can exist at industrial sites. Levels selected
for temperature and relative humidity are not extreme conditions and are
frequently recorded. In this experimental model, the factors are thus fixed.
13
-------
Table 7. ENVIRONMENTAL FACTORS AND LEVELS USED
IN THE CHAMBER EXPOSURE EXPERIMENT
Environmental factors
Sulfur dioxide, yg/m3
Nitrogen dioxide, yg/m3
Ozone, yg/m3
Temperature, C°
Relative humidity, %
Levels
Low
79
94
157
13
50
High
1310
940
980
35
90
In addition to the environmental factors, the total time of exposure and
the time frame of each dew/light cycle were constant for each environmental
condition. Total exposure duration was 1000 hours; however, materials were
also evaluated after exposure periods- of 250 and 500 hours. Each dew/light
cycle lasted 40 minutes and consisted of 20 minutes of darkness and subsequent
formation of dew, and 20 minutes of simulated sunlight.
Since there are five factors at two levels, the complete experimental
design called for 32 (25) difference environmental exposure conditions--one
at each combination of factor and level (Table 8)--and did not include
replication of each exposure condition. The large number of exposure
conditions was necessary because of the many direct and synergistic or
interaction effects that may occur. On the other hand, the probability that
all of these effects would actually occur was quite small. Since each exposure
condition was to be 1000 hours, it was more practical to perform fewer exposures.
Therefore, 16 exposure conditions (represented by Xi through X^ 5 in Table 8)
were selected initially. These exposure conditions included the direct and
interaction effects of the three gaseous pollutants and relative humidity
at the high temperature level. The high temperature level was selected because,
according to reaction kinetics, environmental factors that are statistically
insignificant at high temperatures will most likely show insignificant inter-
actions with temperature. Thus, the results of the first 16 high temperature
exposures would determine which, if any, of the low temperature exposures
(Xj7 - X32) should be conducted. Furthermore, the chances were good that the
low temperature exposures would not be necessary to identify environmental
factors that cause significant effects. For similar reasons, the sequential
performance of exposures was structured rather than randomized. (If effects
could not be observed at high levels, there would be no need to study low levels.)
Possible bias introduced by this procedure was minimized by the type of F-test
used in the analysis of data.
14
-------
Table 8. TWO-LEVEL FACTORIAL ARRANGEMENT
High relative humidity
High
Temperatur
Xi
X3
X5
X7
' X9
Xn
Xl3
Xis
Low
i Temperatun
X17
X19
X21
X23
X25
X27
X29
X31
Low relative humidity
High
Temperatur
X2
X*
X6
Xe
XJQ
Xl2
Xji,
Xl6
Low
Temperature
*
Xj.8
X2o
X22
X24
X26
' x28
X30
XS2
Exposure condition
Low S02
High S02
Low S02
High S02
Low S02
High S02
Low S02
High S02
Low 03
High 03
Low 03
High 03
Low N02
High N02
TREATMENT OF DATA
Statistical analysis of variance was applied to the exposure data for
each material. With the two-level factorial experiment, it was possible to
determine which of the 15 direct and interaction factors had a statistically
significant effect on selected material properties (corrosion rates, erosion
rates, color change, etc.)- Since each exposure condition was not replicated,
the inability to precisely control the factors may have biased the analysis of
variance because it could not be included in the calculation of error vari-
ability. The error term included only within-chamber-variability caused by
position and time effects and possibly non-linearity of the time-damage
function.
If a factor (one of 15 direct or interaction factors) actually has no
effect on a material, the calculated variability associated with that factor
is caused by error. In non-replicated statistically designed experiments,
it is often assumed that the highest order of interaction is the least likely
15
-------
factor to cause an effect and the variability associated with it is used to
calculate the error mean square. Risks, however, are involved in making this
assumption. First, interactions may actually affect the response and, there-
fore, the resulting F test on other factors would be less sensitive. Second,
in a two-level factorial arrangement only one degree of freedom is associated
with an interaction factor; this renders the F test less sensitive than when
using an error term with more degrees of freedom. With a less sensitive F test,
error is.more likely to mask the actual effects of tested factors.
The sensitivity of the F test can be improved by including in the error
term other interaction factors that are not likely to have an effect on the
response. Nevertheless, some degree of risk does exist, since the included
factors may cause effects. The risk, however, is somewhat mitigated by
averaging any effects with the error variability and by the increased number
of degrees of freedom.
When applying the F test for significance in the analysis of variance of
this study, the residual variability associated with all three and higher
variable interaction factors was used to calculate the error mean square.
16
-------
SECTION VII
MATERIALS TESTING TECHNIQUES AND EVALUATION
TESTING TECHNIQUES
Testing techniques to measure effects were selected to provide damage
data that would reflect loss of service life. Table 9 summarizes the techniques
used for each material. Later chapters, which present and discuss the exposure
results for individual materials, give additional details on testing techniques.
Table 9. MATERIALS TESTING TECHNIQUES
Material Testing technique
Weathering steel Gravimetric (loss-in-weight).
Galvanized steel Gravimetric (loss-in-weight).
Aluminum alloy Time to failure (cracking) of stressed
C-rings.
Paint Gravimetric (loss-in-weight). Loss-
in-film-thickness.
Dyed fabric Color change. Microscopic exami-
nation of fibers.
Rubber tire Loss-in-breaking-strength of rayon
cord reinforced white sidewall rubber.
Length and depth of rubber cracking.
Plastic house siding Visual appearance—cracks, discolora-
tion, staining, etc.
Marble and cement Gravimetric (loss-in-weight).
Microscopic examination.
EVALUATION
Before conducting the exposure experiment, it was necessary to evaluate
the techniques for measuring damage to materials and the effectiveness of the
dew/light cycle in accelerating the damage. To assess these parameters,
specimens of the selected materials were exposed for 1000 hours in two chambers
receiving clean air at 35°C and 90 percent relative humidity. One chamber
operated continuously on a 40-minute dew/light cycle; the other operated at
constant temperature and relative humidity (no dew/light cycle).
17
-------
The galvanized steel and paint panels were evaluated for loss-in-film-
thickness after exposure periods of 250, 500, and 1000 hours. The method
of least squares was used to statistically analyze the rate of loss-in-film-
thickness for these materials Tables 10 and 11 present the results from the
two chamber exposure conditions.
Table 10. RATE Op LOSS-IN-FILM-THICKNESS DURING
THE DEW/LIGHT .CYCLE EXPOSURE CONDITION
Material
Number of
observations
Rate of loss-
in-film-thickness t
ym/yr
95% conf.
limits, ym/yr
Galvanized steel
Acrylic coil
coating
Vinyl coil
coating
Oil base paint
Latex paint
36
90
90.
36
36
0
6.13
4.15
267.
25.9
+ 1.49
+ 3.03
+_ 484
+_ 39.0
Table 11. RATE OF LOSS-IN-FILM-THICKNESS DURING THE
CONSTANT TEMPERATURE AND RELATIVE HUMIDITY EXPOSURE CONDITION
Material
Number of
observations
Rate of loss-
in-fi 1m-thickness
ym/yr
95% conf.
limits, ym/yr
Galvanized steel
Acrylic coil
coating
Vinyl coil
coating
Oil base paint
Latex paint
36
90
90
36
36
0
10.37
6.15
32.9
0
+_ 0.56
+_ 1.91
+ 49.4
18
-------
After a small initial loss, the galvanized steel panels exposed in both
chambers had zero loss rates. An initial zinc loss of 4.40 +_ 1.30 ym was
recorded for the dew/light cycle condition, whereas for the constant temperature-
relative humidity condition, 0.42 +_ .93 ym was initially lost. For the exposed
paint specimens, the dew/light cycle accelerated the rate of loss-in-film-
thickness for the oil base and latex house paints, but not the two coil coat-
ings. This anomaly may have been caused by the lack of experimental replica-
tion; however, it was more likely caused by inaccuracy of the measurement
technique. Film thickness was measured using a Dermitron instrument, an
electronic non-destructive thickness tester engineered to produce direct
thickness reading of metallic and non-conductive coatings on steel and non-
magnetic metals. Gravimetric analysis was later used in conjunction with film
thickness measurements.
Tensile strengths (psi) of the rayon cord reinforced rubber tire specimens
were measured after the 1000-hr exposure period for both environmental conditions,
An analysis of variance (Table 12) was made on the breaking strength data to
determine significant effects. The calculated F statistic revealed that
Table 12. ANALYSIS OF VARIANCE FOR RAYON
TIRE CORD BREAKING STRENGTH
Chamber
Stress**
Chamber x
Stress
Error***
Total
Sum of
squares
9894.4
3369.8
965.7
3639.5
17869.4
Degrees of
freedom
1
1
1
8
11
Mean squares
9894.4
3969.8
965.7
454.9
F F
calc. table
21.75* 5.32
7.41 5.32
2.12 5.32
* Exceeds the 5 percent probability value
**Rubber strips were stretched 10 and 20 percent
***
Replication error
the chamber and stress effects were statistically significant at the 5 percent
probability level. However, the interaction effect of chamber x stress was
insignificant. In contrast to these results, no significant difference was
found in the breaking strength of the white sidewall rubber (after the
19
-------
reinforcing cord had broken) exposed to the two chamber conditions. Since the
exposure conditions affected the cord strength but not the rubber, and.since
the rubber protected the cord from damage by light from the xenon lamp, the
cord effect must have been a result pf differences in moisture content. The
strength of rayon generally decreases with an increase in moisture content.
By means of absorption and wicking action, moisture entered the exposed ends
of the rayon cord in the tire specimens. To prevent this action in subse-
quent exposures, the ends of rubber tire specimens were dipped in wax.
The dew/light cycle accelerated the weathering of the remaining exposed
materials. Measurement techniques for these materials provided damage data
for which significant differences were distinguished. As a result, no
additional techniques were needed to assess damage of these materials.
20
-------
SECTION VIII
EFFECTS ON WEATHERING STEEL
Weathering steel was developed for use in unpainted exterior architectual
designs. In most environments, the steel develops a small grained strongly
adherent rust coating that tends to serve as a protection film against subse-
quent corrosion. Weathering steel has been used in the design and construction
of the Chicago Civic Center, the Washington Battelle Memorial Institute Building,
and the New York Ford Foundation Office.
METHODOLOGY
The weathering steel exposed in this study was Cor-Ten A, a product of
United States Steel Corporation. Table 13 gives the elemental analysis of
Cor-Ten A. The measured yield point and ultimate tensile strength were 35,820
N/cm2 and 54,480 N/cm2, respectively.
Table 13. ELEMENTAL ANALYSIS OF WEATHERING STEEL
Elemental Composition, %
C
0.11
Mn
0.37
P
0.10
S
0.018
Si
0.39
Cu
0.29
Ni
0.11
Cr
0.90
Fe
Remainder
Corrosion rates of most metals normally accelerate when metals are exposed
to polluted environments. The rate of corrosion, therefore, was selected as the
means for assessing the effects produced by the various controlled environment
exposure conditions. Corrosion was measured by the weight-loss method. This
method consisted of weighing the panel before exposure, and removing the
corrosion products and reweighing after exposure. The difference in weight
was a measure of the amount of metal that corroded during the exposure period.
Panels measuring 7.6 cm x 12.7 cm were sheared from sheet stock. The
panels were cleaned by scrubbing them in a warm detergent solution followed
by separate rinses in distilled water and acetone. They were then weighed
on an analytical balance to within +^0.1 mg.
Six panels were exposed to each of the 16 polluted and 4 clean air
conditions. The panels were randomly placed on chamber exposure racks and two
panels were randomly removed after nominal exposure periods of 250, 500, and
1000 hours.
21
-------
Corrosion products on the exposed panels were chemically removed by
immersion for one hour or longer in Clark's solution5, followed by scrubbing
and final rinses in distilled water and acetone. The panels were then weighed,
the weights corrected for the loss of base metal [0.06 g/ (panel hr.)] while
in Clark's solution, and weight-loss values calculated. To render the weight-
loss values (g/panel) more meaningful, they were converted to equivalent
thickness-loss values (ym) by using the factor 13.345 ym/g.
RESULTS AND DISCUSSION
Corrosion of weathering steel is normally a parabolic function of time.
The corrosion data collected in this study, however, were essentially linear
with time. The relatively short exposure periods (1000 hrs.) probably
accounted for the deviation.
Because of the linear relationship, a mean corrosion rate for each ex-
posure condition was calculated using the method of least squares through
the origin. Tables 14 and 15 present individual corrosion rates for the
polluted air and clean air exposures, respectively. Corrosion rates in
polluted air ranged from 84 to 762 ym/yr with an average of 383 ym/yr. Best
estimate of the standard deviation on individual corrosion rates in polluted
air was +_ 72 ym/yr. Clean air corrosion rates ranged from 1 to 86 ym/yr.
The highest clean air corrosion rate was about the same as the lowest rate
in polluted air.
Table 14. CORROSION RATES OF WEATHERING STEEL
EXPOSED TO DESIGNATED CONTROLLED POLLUTED AIR
ENVIRONMENTAL CONDITIONS
Corrosion rate and standard deviation, vm/yr
High relative humidity
High S02
753 + 85
762 + 121
736 + 90
656 + 56
Low S02
256 i 27
178 +44
230 + 27
147 + 13
Low relative humidity
High S02
414 + 151 •
607 + 105
479 + 78
371 i 18
Low S02
179 i 30
162 + 62
123 + 26
84 + 14
Exposure
condition
High 03
Low 03
High 03
Low 03
High N02
Low N02
Note: Corrosion rates based on six data sets per exposure condition.
22
-------
Table 15. CORROSION RATES OF WEATHERING STEEL
EXPOSED TO DESIGNATED CONTROLLED CLEAN AIR
ENVIRONMENTAL CONDITIONS
Corrosion rate and
standard deviation, um/yr
High relative
humidity
86 + 31
1.03 + 0.17
Low relative
humidity
28 + 10
1.07 + 0.06
Temperature
35°C
13°C
Note: Corrosion rates based on six data sets per
expos ure cond i ti on.
Tables 16 and 17 show analyses of variance of the experimental data for
the polluted and clean air exposure conditions, respectively. On the basis of
calculated F values, the important factors controlling corrosion were concen-
tration of SC>2, relative humidity, and temperature. The interaction between
S02 and relative humidity was significant at a lower confidence level.
As shown in previous studies6*7, electrolytic corrosion occurs only
when metal panels are Vet. During dew/light cycles in the chambers, panel
temperatures continually varied, thus causing changes in relative humidity
adjacent to the panels. When temperatures were below the dew point, moisture
condensed on the panels. Because of the hygroscopic nature.of corrosion
products, panels also became moist even at temperatures above the dew point.
Thermocouples attached to panels were used to monitor panel temperatures
during the dew/light cycles for each clean air exposure condition. Figure 2
shows the results. Theoretically, the critical factors controlling corrosion
are the time-of-panel-wetness and the geometric mean temperature during the
time panels are wet. Since the moisture content of the air entering the
chambers was essentially constant (set by the input conditions), time-of-panel-
wetness per cycle was estimated by using the cycling temperature data to
calculate relative humidity values (adjacent to panels) as a function of time.
Visual observations, however, detected moisture on panels at a calculated
relative humidity of about 85 percent. This was probably due to the hygro-
scopic nature of the corrosion products. Thus, panels were considered wet
at temperatures corresponding to relative humidities equal to or greater than
85 percent. Table 18 shows the calculated times-of-panel-wetness per dew/light
23
-------
Table 16. ANALYSIS OF VARIANCE OF WEATHERING STEEL CORROSION
DATA FOR POLLUTED AIR EXPOSURE CONDITIONS
Factor
S02
RH
03
N02
S02 x RH
S02 x N02
S02 x N02
RH x 03
RH x N02
03 x N02
Residual
Total
Contrast
213.7
12.7
12.7-
30.4
48.3
-14.4
6.5
16.3
-7.8
-26.1
Sum of
squares
730726
105138
2590
14810
37264
3322
673
4231
975
10870
14155
924754
Degrees
of freedom
1
1
1
1
1
1
1
1
1
1
5
15
Mean
square
730726
105138
2590
14810
37264
3322
673
4231
975
10870
2831
F
258.11**
37.14**
0.91
5.23
13.16*
1.17
0.24
1.49
0.34
3.84
.
R2
0.790
0.114
0.003
0.016
0.040
0.004
. 0.001
0.005
0.001
0.012
Q.003
Notes: F uses the residual mean square in the denominator
R2 is the coefficient of determination excluding the within chamber variability
** 99 percent probability level of significance
* 95 percent probability level of significance
Table 17. ANALYSIS OF VARIANCE OF WEATHERING STEEL CORROSION DATA
FOR CLEAN AIR EXPOSURE CONDITIONS
1
Factor
RH
Temperature
RH x T
Error
Total
Contrast
14.49
27.75
14.51
Sum of
squares
839.8404
3080.2550
842.1604
5181.7875
9944.0383
Degrees
of freedom
1
1
1
20
Mean
square •
839.8404
3080.2500
842.1604
259.0894
F
3.24
11.89**
3.25
R2
0.085
0.310
0.085
0.521
1
** 99 percent probability level of significance
24
-------
to
INPUT RELATIVE HUMIDITY %'.',
I ! I I ! I
20
30 40 0
CYCLE TIME-MINUTES
INPUT RELATIVE HUMIDITY 50",
Figure 2. Temperature-time profiles for exposed metal panels during chamber
dew/light cycles.
-------
Table 18. TIME-OF-PANEL-WETNESS PER DEW/LIGHT CYCLE
AND GEOMETRIC MEAN PANEL TEMPERATURE WHEN WET AS A
FUNCTION OF INPUT TEMPERATURE AND RELATIVE HUMIDITY
Input
temperature
35°C
13°C
Input relative humidity
90%
24.5 min
29.4°C
22.0 min
10.2°C
50%
20.7 min
20.4°C
7.0 min
4.6°C
Note:. Calculated values are for both weathering steel and galvanized
steel.
cycle and geometric mean temperatures of panels when wet for each clean air
exposure cpndition.
Because steel corrodes only when wet, the corrosion data were analyzed in
terms of the expected times-of-wetness. On the basis of long term atmospheric
corrosion studies8'9, the corrosion-time function should be more nearly para-
bolic than linear as was observed in the short term chamber exposures. In the
clean air exposures, the reactants (steel, water, and oxygen) were present
in excess and the reaction rate should be activation controlled. Thus, the
theoretically expected relationship between corrosion, time, and temperature
should be:
E
a° " If
where:
Corr
tw
E
R
T
a.
thickness-loss, urn
time-of-wetness, yr
activation energy, cal/(g-mol °K)
gas constant, 1.9872 cal/g-mol
geometric mean temperature of panels when wet,
regression coefficient
26
-------
A least squares fit of the clean air experimental data resulted in
(2)
or
-
Corr = t e • RT (3)
This empirical function accounted for 92 percent of the variability in the
clean air experimental data.
The reaction of steel with moisture and oxygen occurs with or without
862 present. Thus, the contribution of moisture and oxygen to the total
corrosion may be excluded when analyzing the direct and synergistic effects of
S02- This amount of corrosion, therefore, was calculated and substracted from
each corresponding data set from the polluted air exposures and the resulting
data fitted to the relationship:
= ^ ^ + ^ £R ^ _ E_ ^
where: Corr = thickness-loss, ym
t = time-of-panel-wetness, yr
S02 = concentration of S02, yg/m3
E = activation energy, cal/(g-mol °K)
R = gas constant, 1.9872 cal/g-mol
T = geometric mean panel temperature when wet, °K
a0, a. = regression constants
E
The calculated constant, 5-, was small, positive, and statistically in-
significant. This fact strongly suggested that the reaction of S02 with steel
was diffusion controlled rather than activation controlled. The coefficient
for £n S0£ was not statistically different from 0.5; corrosion, therefore,
appeared to be proportional to /S02. Thus, the weathering steel experimental
data were re-evaluated as a function of the square root of the dose, /twS02,
and the resulting best fit of the theoretically consistent relationship was:
,,.,- ,, 31150,
Corr = [5.64 SSO^ + e &*-™ RT~J]^~ (5)
27
-------
This empirical function accounted for 91 percent of the variability in the
experimental data from both the clean and polluted air exposure conditions.
Figure 3 shows a plot of predicted versus experimental data. The devi-
ation at the high end of the graph is probably because the experimental data
were more consistent with a linear-time function than a parabolic-time function.
A better test of an empirical function is to determine how accurately it
predicts actual long term atmospheric corrosion results. Fortunately, corrosion
data for a similar weathering steel are available from a field study conducted
in eight cities.8 Although time-of-panel-wetness and temperature during time-of-
wetness were not measured in the field study, average temperature and relative
humidity during the total time of exposure were calculated. Thus, if relationships
between average relative humidity and time-of-panel-wetness, and between average
temperature and geometric mean temperature when panels are wet can be obtained,
the empirical function may be expressed in terms suitable for making comparisons
with the field data.
Guttman7 has plotted the probable time-of-wetness as a function of rela-
tive humidity for both groundward and skyward surfaces of exposed metal panels.
Assuming the average probability for the two surfaces represents the fraction
of the total time panels are wet, the data fit a relationship
fw = e 4'04 - TOT (6)
where: f = fractional time-of-panel-wetness
RH = relative humidity, %
On the basis of the temperature cycle data in Figure 2, the geometric
mean panel temperature when wet was about 6-°K less than the average panel
temperature for the total time of exposure. This information together with
the field study data, were used to calculate factors (Table 19) that affected
the atmospheric corrosion of weathering steel in each city. Thus, these
factors may be used in the empirical function to predict the amount of corrosion
likely to develop in each city.
A measure of corrosion predictability is the coefficient of variation,
R2, between predicted and field study values. Values of R2 were calculated
for each city of the field study; results are given in Table 20. With the
exception of Los Angeles, predicted corrosion values agreed poorly with
actual corrosion measured in cities with relatively high oxidant levels. The
Los Angeles data was probably in good agreement because the low time-of panel-
wetness did not allow much corrosion to develop, thus the magnitude of error
was small. Regressions on the field data, however, strongly suggested that
oxidants inhibited corrosion. The major oxidant in natural environments is
ozone, but the chamber exposures showed that ozone produced neither inhibiting
28
-------
oo
o
_J
00
UJ
z
tJ
H—I
3:
I—
Q
LU
OH
TOO
90
80
70
60
50
40
30
20
10
10 20 30 40 50 60 70
PREDICTED THICKNESS-LOSS,
80 90
100
Figure 3. Comparison of predicted and measured thickness-loss
values for weathering steel exposed to laboratory controlled
polluted air and clean air conditions.
29
-------
Table 19. FACTORS AFFECTING ATMOSPHERIC CORROSION OF
WEATHERING STEEL EXPOSED AT URBAN SITES
Geometric mean
panel tempera-
ture when wet,
City °K
Chicago
Cincinnati
Detroit
Los Angeles
New Orleans
Philadelphia
San Francisco
Washington
* [5.64 /S"U7 +
277.3
280.2
277.2
285.2
287.5
279.1
281.3
281.0
(55.44-31,
[
Fractional
time-of-
panel -wet-
ness, fw
0.1136
0.1628
0.1367'
0.0841
0.2992
0.1248
0.2244
0.0932
,150)! n —
It J Tw
Average
pollutant
concentra-
tion, yg/m3
S02
406
79
118
39
24
218
34
126
03
47
59
24
75
35
57
37
50
Combined
factor*
38.42
20.47
22.77
10.68
16.48
29.60
15.94
19.54
30
-------
Table 20. CORROSION PREDICTABILITY AS MEASURED BY
COEFFICIENTS OF VARIATION, R2, BETWEEN
PREDICTED AND FIELD VALUES
Coefficient of variation, R2,
City a measure of corrosion predictability
Chicago 0.976
Cincinnati 0.544
Detroit 0.997
Los Angeles 0.984
New Orleans 0.998
Philadelphia 0.392
San Francisco 0.978
Washington 0.701
Overall 0.846
nor accelerating effects. In the field study, some other oxidant, or unmeasured
factor that was covariant with oxidant, apparently caused the inhibiting
effect. Excluding the data from the high oxidant cities, the average co-
efficient of variation was 0.986. This was exceptionally good, expecially
considering the simplifying assumptions that were made.
31
-------
SECTION IX
EFFECTS ON GALVANIZED STEEL
Galvanized steel is widely used for such exterior applications as roofing,
guttering, highway railing, and chain link fencing.
METHODOLOGY
Commercial grade 18 gauge galvanized steel sheet with an approximate 25
ym zinc coating was selected. The effects assessment method (weight-loss),
panel size, pre-exposure cleaning procedure, and panel exposure sequence were
the same as those used for the weathering steel.
Corrosion products on the exposed galvanized steel panels were chemically
removed by immersion for about 10 minutes in a 10 percent aqueous ammonium
chloride solution maintained at 72 to 80°C. The correction factor for zinc
lost during chemical cleaning was 0.009 g/(panel rain). Multiplying the
corrected weight-loss values (g/panel) by 14.98 ym/g converted them to equiv-
alent zinc film thickness-loss values (ym).
RESULTS AND DISCUSSION
t
As was theoretically expected, corrosion of the galvanized steel was
essentially linear with time. Thus, an average corrosion rate for each of
the 16 polluted air exposure conditions was calculated by the method of least
squares through the origin. Table 21 shows the corrosion rates; they ranged
from 3 to 33 ym/yr, and the best estimate of the standard deviation on
individual rates was +_ 5 ym/yr.
Corrosion rates for galvanized steel exposed to clean air conditions were
unexpectedly high and have been presented and discussed as a separate publi-
cation. 10
An analysis of variance (Table 22) was performed on the calculated
polluted air corrosion rates. Both sulfur dioxide (S02) and relative humidity
produced highly significant positive effects (99 percent probability level)
on the corrosion rate of galvanized steel. Excluding within chamber error,
these two factors accounted for 94 percent of the variability. Ozone also
appeared to be a significant factor at the 95 percent probability level;
however, it only accounted for about 3 percent of the variability. Therefore,
it was not considered in further analysis of the data.
According to accepted corrosion theory,.zinc surfaces corrode only when
wet. Therefore, the data in table 21 may be reevaluated in terms of time-of-
32
-------
Table 21. CORROSION RATES OF GALVANIZED STEEL
EXPOSED TO DESIGNATED CONTROLLED POLLUTED AIR
ENVIRONMENTAL CONDITIONS
Corrosion rate and standard deviation, pm/yr
High relative humidity
High S02
33 + 7.1
28 + 1.5
29 + 10
25 +.3.1
Low S02
17 + 5.2
12 +4.3
20 + 11
16 +6.6
Low relative humidity
High S02
19 +4.1
15 + 2.1
17 +1.3
16 +_3.5
Low S02
6.6 +0.9
3.8 +0.6
4.0+0.9
3.2 +2.1
Exposure
condition
High 03
Low 03
High 03
Low 03
High N02
Low N02
Note: Corrosion rates based on six data sets per exposure condition.
Table 22. ANALYSIS OF VARIANCE OF GALVANIZED STEEL
CORROSION DATA FOR POLLUTED AIR EXPOSURE CONDITIONS
Factor
S02
RH
03
N02
S02 x RH
S02 x 03
S02 x N02
RH x 03
RH x N02
03 x N02
Residual
T6tal
Contrast
6.2188
6.0272
1.5893
-0.2644
0.0565
0.1374
-0.8088
0.5486
0.3598
-0.4093
Sum of
squares
618.7781.
581.2318
40.4146
1.1188
0.0512
0.3022
10.4668
4.8147
2.0714
2.6806
16.3176
1278.2478
Degrees
of freedom
1
1
1
1
1
1
1
1
1
1
5
15
Mean
square
618.7781
581.2318
40.4146
1.1181
0.0512
0.3022
10.4668
4.8147
2.0714
2.6806 .
3.2635
F
189.61**
178.10**
12.38*
0.34
0.02
0.09
3.21
1.48
0.63
0.82
R2
0.484
0.455
0.032
0.000
0.000
0.000
0.008
0.004
0.002
0.002
.0.013
Notes: F uses the residual mean square in the denominator
R2 is the coefficient of determination excluding the within chamber variability
** 99 percent probability level of significance
* 95 percent probability level of significance
33
-------
panel-wetness and geometric mean panel temperature (Table 18) rather than total
time. To convert to corrosion rates during time-of-panel-wetness, the corrosion
results in Table 21 were divided by the time-of-wetness expressed as a fraction
(0.6125 for high relative humidity and 0.5175 for low relative humidity).
The corrosion of galvanized steel should be consistent with the following
theoretical relationship:
Corr = (a0 S02 + e b " E/RT) tw (7)
where: Corr = zinc thickness-loss, ym
a0,b = regression coefficients
562 = SC>2 concentration, yg/m3
tw = time-of-wetness, yr
E = activation energy, cal/(g-mol °K)
R = gas constant, 1.9872 cal/g-mol
T = geometric mean temperature of panels when wet, °K
A best fit of the experimental data yielded:
41.85 - 23240/RT
Corr = (0.0187 S02 + e ) tw (8)
This empirical function accounted for 91 percent of the variability in the
experimental data from the polluted air exposure conditions. A comparison
of the measured zinc thickness-loss values with predicted values is shown
in Figure 4.
Previous research has shown that zinc corrodes more rapidly in urban
and industrial areas than in rural areas.11'12 Researchers suspect that
atmospheric pollutants, especially sulfur dioxide, accelerate the corrosion
of zinc by preventing the formation of a protective film, which probably
consists of an insoluble zinc oxide and carbonate mixture. Several investi-
gators have developed relationships between the amount of zinc corrosion and
atmospheric concentration of sulfur dioxide.7>13 The previous research,
however, involved field exposures of high purity zinc panels to environmental
factors that could not be controlled.
In the chamber study, galvanized steel panels were exposed to controlled
levels of gaseous pollutants, temperature, and relative humidity. No attempt
was made to remove oxygen or carbon dioxide from the air circulated through
the chambers. These gases along with the pollutants were, therefore, readily
available to react with the zinc coating during the time-of-panel-wetness.
34
-------
O
_l
I
(/}
oo
Q
UJ
PREDICTED THICKNESS-LOSS, ym
Figure 4. Comparison of predicted and measured thickness-loss
values for galvanized steel exposed to laboratory controlled
polluted air.
35
-------
Two reactions are believed to occur simultaneously in the presence of the
chamber pollutants: (1) the formation of zinc hydroxide or oxide, and (2) the
formation of zinc sulfate. The analysis of variance of the chamber data, how-
ever, indicated that of the three gaseous pollutants, only sulfur dioxide
significantly accelerated the corrosion fate. Therefore, the formation of
sulfur containing corrosion products would be expected. Scanning electron
photomicrographs of exposed panels (figure 5) revealed crystalline material
uniformily dispersed over the zinc surface. X-ray diffraction patterns of
the corrosion products were complex and indicated one or more unidentifiable
hydrated compounds. Microprobe scans, however, revealed sulfur atoms, likely
in the form of sulfate, to be uniformily distributed over the zinc surface.
During the dew cycle, the pH of the thin layer of condensate that formed
on the galvanized panels was measured and found to remain acidic (pH of 5.6).
However, the zinc corrosion products (oxides or hydroxides and sulfates) are
known to be soluble in acidic media. Corrosion rates, therefore, should remain
constant with time because the acidic dew not only dissolves the corrosion
products but distributes them uniformily across the zinc surface (figure 5).
The process continues during each cycle of dew formation.
The reactant (oxygen) producing the hydroxide or oxide was readily available
during exposure; thus, the reaction with the zinc surface should be activation
controlled. In comparison, sulfur dioxide levels were considerably lower;
therefore, the formation of sulfate should be diffusion controlled. Carbon
dioxide was also readily available to react with the zinc coating; however,
the protective zinc carbonate film was not likely to form in the presence of
an acidic (sulfurous acid) medium.
The best fit of the corrosion data produced an empirical function that
accounted for 91 percent of the variability. Unfortunately, the predictability
of this relationship cannot be tested because no comparable field data exist
for exposed galvanized steel. However, field data for the corrosion of pure
zinc panels are available.13 After making minor tranformations in the empiri-
cal corrosion function, predicted corrosion values for galvanized steel were
consistently lower than corrosion values for pure zinc measured during field
exposures. The empirical function accounted for only 29 percent of the
variability in the field data.
36
-------
Figure 5. Scanning electron photomicrograph of an
exposed galvanized steel panel showing uniformily
dispersed crystalline material over the zinc surface,
37
-------
SECTION X
EFFECTS ON ALUMINUM ALLOY
A number of catastrophic failures of metal structures, such as bridges,
towers, and aircraft components, have been caused by a phenomenon called
stress corrosion cracking. Broadly speaking stress corrosion cracking is a
form of localized failure that is more severe under the combined action of
stress and corrosion than would be expected from the sum of the individual
effects of stress and corrosion acting alone. Failure is characterized by
a brittle-type fracture in an otherwise ductile material. In some cases, the
cause of failure has been linked to air pollution.
Several important alloys are susceptible to stress corrosion cracking.
The most familiar example is brass which fails in environments containing
ammonia. Investigators have found that a number of aluminum alloys also fail
by stress corrosion cracking especially in seacoast environments. However,
the 7000 series of high strength aluminum alloys, which contain little or no
copper, is more susceptible to stress corrosion cracking in industrial environments
than along the seacoast. Since air pollution is synonymous with industrial
environments and is therefore a likely cause of failure, an alloy from this
series was selected as a representative metal for assessing stress corrosion
cracking.
METHODOLOGY
High strength aluminum alloy 7005-T53 was chosen for exposure. Table
23 gives the chemical analysis of this alloy. The main reason for choosing
Table 23. ELEMENTAL ANALYSIS OF
ALUMINUM ALLOY 7005-T53
Elemental Composition, %
Si Fe Cu Mn Mg Cr Ni Zn Ti Zr Al
0.09 0.22 0.04 0.40 1.58 0.09 <0.01 4.61 0.03 0.13 Remainder
this particular 7000 series alloy was that it was readily available as
extruded tubing. This shape made it possible to easily fabricate C-ring
test specimens, one of the common types of specimens used to assess stress
corrosion.
38
-------
The tubing had an outside diameter of about 5.7 cm with a wall thickness
of about 0.23 cm. Fabricated C-ring specimens were 1.9 cm wide with machined
edges; inside and outside surfaces remained as received.
C-ring specimens were stressed according to ASTM recommendations;11* actual
dimensions were measured to within 10 ynr. Two levels of stress were used:
2.07 x 108 N/m3 (30 ksi) and 2.76 x 108 N/m3 (40 ksi). Figure 6 shows the design
of a stressed C-ring specimen. To measure time-of-failure, each specimen was
Figure 6. Method of stressing C-ring specimens
spring loaded and electrically connected to a recorder. An insulated wire was
attached across the compressed spring. Failure of a C-ring specimen caused
the wire to break, thus signaling the time-of-failure.
To simplify the electrical wiring scheme, a group of three equally stressed
C-ring specimens were mounted on a metal (aluminum alloy 6061-T6) plate that
was then placed at random on a chamber exposure rack. For each exposure condition,
two groups (one at each stress level) of three specimens were exposed.
Each group of three specimens was wired according to Figure 7 so that
time-of-failure for each specimen could be determined by recording step changes
in voltages on a ten channel multipoint recorder. The resistance R in figure
7 was varied from one group to another to separate voltage traces on the recorder.
Resistance values ranged from 0 to 35 ft. When a specimen breaks, the recorded
voltage increases and thus marks the time-of-failure.
39
-------
4.5 v
150Kfl
15 0
150
C-RINGS
Figure 7. Wiring of specimens to record times-of-failure
RESULTS AND DISCUSSION
None of the specimens exposed for 1000 hours to the 16 polluted air environ-
ment conditions completely failed; thus, time-to-failure could not be used
as a response.
Specimens developed small intergranular cracks perpendicular to the
stress direction (Figure 8). From a visual standpoint, number and lengths
of these cracks appeared to be a function of exposure time, severity of
the environment, and level of stress.
Selected specimens were examined by both light and scanning electron
microscopy. The cracks were intergranular and microprobe analysis revealed
sulfur on the fracture faces (Figure 9).
A 5.1 cm cord across the stressed area was cut from each C-ring specimen
after 1000 hours exposure, and compressed to failure by bending in a tensile-
compression testing instrument. The maximum load required to bend the
specimen was assumed to be a function of the remaining uncracked cross
sectional thickness and was recorded as the response to the controlled
environmental factors. A typical load-deflection curve on these specimens
had two maxima. The first represented buckling and the second bending;
the second maximum was used as the response.
Results for the polluted air exposure conditions are given in Table 24
and ranged from a low of 288 N to a high of 804 N. Low values represent cross
sections reduced by stress corrosion cracks. The best estimate of the standard
deviation on any one set of data was ± 108 N.
40
-------
Figure 8. Stress induced integranular cracks
in aluminum alloy 7005-T53 after 1000 hours ex-
posure to air containing S02. (Mark = 1 mm)
41
-------
SCANNING ELECTRON MICROGRAPH
MICROPROBE X-RAY SCAN FOR SULFUR Ko
Figure 9. Fracture face cross section of aluminum alloy specimen ruptured by
bending after intergranular cracks developed during 1000 hours exposure to air
containing SC>2. (Mark = 100 ym)
-------
Table 24. BENDING STRENGTH OF ALUMINUM ALLOY STRESS CORROSION
SPECIMENS AFTER EXPOSURE TO DESIGNATED CONTROLLED POLLUTED
AIR ENVIRONMENTAL CONDITIONS FOR 1000 HOURS
Stress,
KN/m2
276
207
276
207
276
207
276
207
Bending strength and
standard deviation, N
High relative humidity
High S02
522 +_ 107
693 + 137
288 +, 175
482 + 32
632 + 66
512 + 105
534 + 32
690 + 6
Low S02
524 +_ 173
725 + 31
740 + 72
764 + 69
804 +' 58
791 + 30
486 i 127
712 +_ 71
Low relative humidity
High S02
503 + 249
690 + 148
570 + 118
611 i 115
396 i 189
529 + 34
591 £ 35
669 + 123
Low S02
780 + 185
760 +_ 77
777 + 14
705 + 20
657 i 58
685 + 73
681 + 176
728 + 66
Exposure
condition
High 03
Low 03
High 03
Low 03
High NO,
*. L.
\ nw MO
Note: .Bending strength values based on three data sets per exposure condition.
Bias may have been introduced into this experiment because of the method
used to expose the C-ring specimens. From the standpoint of experimental
design, a better method would have been to expose individual specimens randomly
placed on a chamber rack, rather than to expose three equally stressed specimens
as a group and randomly place the group on the chamber rack. Thus, because
of grouping, a possible position effect could be confounded with some tested
effects. In analyzing variance, the possibility of arriving at erroneous
conclusions of statistical significance for some variables was minimized
by assuming that all triple and higher interaction effects were caused by
error and using the values of those effects to calculate the error mean
square. The resulting F test for significance, therefore, was much more
stringent than would have occurred by assuming that the only error was associated
to within-sample-variance.
43
-------
Using this test, only changing the level of S02 caused a significant
effect on the bending strength at the 99 percent confidence level. Mean
bending strengths and standard deviations on the means were 565 N ± 23 N
and 713 N ± 18 N for S02 levels of 1310 yg/m3 and 79 yg/m3, respectively.
Level of SC>2 accounted for 26 percent of the total variability.
Although stress was necessary to produce cracks, the two levels of
stress used in this experiment did not produce significant differences in
bending strength.
The results of the clean air exposures are given in Table 25. Assuming
that the triple interaction was caused by error, analysis of variance indicated
Table 25. BENDING STRENGTH OF ALUMINUM ALLOY STRESS CORROSION
SPECIMENS AFTER EXPOSURE TO DESIGNATED CONTROLLED CLEAN
AIR ENVIRONMENTAL CONDITIONS FOR 1000 HOURS
Stress,
KN/m^
276
207
Breaking strength and
standard deviation, N
High relative humidity
High
temp.
697 +187
730 + 79
Low
temp.
633 +_ 259
507 + 284
Low relative humidity
High
temp.
700 + 24
822 +_ 87
Low
temp.
882 + 61
770 + 101
Note: Bending strength values based on three data sets
per exposure condition.
44
-------
that none of the effects were statistically significant. The mean bending
strength and standard deviation on the mean were 717 N ± 36 N.
The mean bending strength of unexposed, unstressed control specimens
was 780 N with a standard deviation on the mean of ± 39 N. Thus, stressed
specimens exposed in clean air environments experienced an 8.1 percent loss
in bending strength. Specimens exposed to 79 pg/m3 and 1310 ug/m3 of S02,
respectively, lost approximately 8.6 percent and 27.6 percent of their
bending strength.
The lack of complete failures, the relatively large number of cracks
perpendicular to the applied stress, and the relatively slow crack growth
rates suggest that the effects mechanism is a type of stress-accelerated
intergranular corrosion rather than conventional stress corrosion cracking.
In a later experiment, specimens exposed stressed for over 2000 hours in
clean air at input relative humidity of 80 percent and temperature at 29°C,
failed by the conventional stress corrosion cracking mechanism. In each
of these specimens, one primary crack with a few secondary cracks developed
(Figure 10).
The amount of damage associated with the exposures in polluted air
appeared to be proportional to the dose of SC>2. For the amount of time
these specimens were exposed at two stress levels, the bending strength
appeared to decrease by 1.5 percent for each 100 yg/m3 of S02-
Longer exposure times may have caused complete failures by a stress
corrosion cracking mechanism. Stress-accelerated intergranular corrosion,
although a significant problem, is not nearly as critical to parts failures
as stress corrosion cracking because the progress of the former can be detected
during periodic inspections while the latter cannot.
Longer exposure times at different S02 levels are needed to determine
if S02 significantly promotes stress corrosion cracking.
45
-------
Figure 10. Stress corrosion crack after 2000 hours
exposure to clean air. (Mark = 1 mm)
46
-------
SECTION XI
EFFECTS ON PAINTS
Painted surfaces represent a tremendous area of potential damage by
air pollutants. The surfaces, however, are covered by a wide variety of
paint formulations that must meet different end use requirements. To evaluate
all the various types of paints would be a time consuming and unreasonable
task. Therefore, on the basis of discussions with paint industry representatives,
four classes of paints, all formulated for exterior exposures, were selected
for testing. These were (1) oil base house paint, (2) acrylic latex house
paint, (3) vinyl coil coating, and (4) acrylic coil coating. Coil coating,
which is applied in factories, is an efficient process for continually dip
coating and curing long, flat strips or sheets of metal.
METHODOLOGY
To insure good heat transfer during exposure, 7.6 cm x 12.7 cm aluminum
panels served as the substrate for the four types of paint. Vinyl and acrylic
coil coating panels were cut from commercially available rolled stock.
The two house paints were sprayed on both sides of unprimed panels, which
previously had been scrubbed in soapy water, rinsed in distilled water and
acetone, and etched with 10% acetic acid. Table 26 shows the nominal paint
film thickness on panels for each of the four types of paint. All of the
coated panels were conditioned for 48 hours in a constant temperature and
relative humidity cabinet at 25°C and 45 percent relative humidity prior
to making initial measurements and after each exposure period. This procedure
Table 26. NOMINAL PAINT FILM THICKNESS ON EXPOSED PANELS
Type of paint Thickness, ym
Oil base house paint* 58
Acrylic latex house paint* 43
Vinyl coil coating 27
Acrylic coil coating 20
''Film allowed to age for three weeks at ambient conditions
allowed each coating to equilibrate, thus minimizing differences in moisture
content of the films.
47
-------
Three panels of each coating were randomly mounted on a chamber exposure
rack. After exposure periods of 250, 500, and 1000 hours, the panels were
removed from the chambers, cleaned, equilibrated, measured for changes, and,
except for final exposure period, returned for additional exposure. The
cleaning procedure consisted of wiping off the silicone paste, washing in 5
percent detergent solution followed by a final rinse in distilled water.
After measurements were taken, silicone paste was reapplied and the panels
were remounted on the rack for the next exposure period.
Two types of measurements were made to assess the amount of damage caused
by exposure in the chambers: (1) loss-in-weight and (2) loss-in-film-thickness.
Panels were weighed on an analytical balance to within ^0.1 mg; film thickness
was measured to within + 0.1 ym using a Dermitron thickness tester. Average
film thickness on each panel was based on 10 measurements (two measurements
at 5 different locations).
Using these measurements, film erosion rates (loss-in-film-thickness per
unit of time) were calculated for each of the four paints. Erosion rates
are a more appropriate means to assess the effects of pollutants on the useful
life of various coatings because paint manufacturers formulate coatings to
fail by gradual erosion of the film during exposure. The erosion results for
each paint, however, should not be misconstrued as representative of an entire
class of paint. Coating formulations may vary considerably within a class of
paint. Formulation as well as exposure condition, therefore, will determine
erosion characteristics. For example, atmospheric sulfur dioxide will affect
paints formulated with calcium carbonate.
RESULTS AND DISCUSSION
The exposure results for the acrylic latex house paint were invalid; this
problem will be discussed later. For the remaining paints, however, the
various controlled exposure conditions yielded significantly different loss-in-
weight values. Furthermore, a definite correlation between loss-in-weight and
loss-in-film-thickness was found for each paint. Because units of film thick-
ness rather than weight are more useful in estimating paint life, and because
weight measurements were more accurate than film thickness measurements, a
conversion factor based on experimental data from initial exposure conditions
was calculated for each paint. The conversion factor is the slope of a least
squares fit of the loss-in-film-thickness versus loss-in-weight data, which
intercepts at the origin. Table 27 gives the resulting factors. Once the
conversion factors were determined, film thickness was no longer measured on
panels subjected to the remaining exposure conditions. Thus, each loss-in-weight
measurement was converted to an equivalent loss-in-film-thickness value prior
to data analysis.
-------
Table 27. FACTORS FOR CONVERTING LOSS-IN-WEIGHT TO
LOSS-IN-FILM-THICKNESS
Type of paint Conversion factor,
ym/g
Oil base house paint 38.53
Vinyl coil coating 17.94
Acrylic coil coating 34.88
Q
Loss-in-weight data for acrylic latex house paint were invalid.
Previous field exposure studies15 indicate that paint film erosion rates
are normally linear after nine months of exposure; in laboratory studies
linearity generally begins after several hundred hours of exposure. Therefore,
a linear least squares fit of loss-in-film-thickness values versus exposure
time was calculated for each paint for each exposure condition using the
following relationship:
y = A + B t (9)
where
p = loss-in-film-thickness, ym
t = time, yr
A = Constant (intercept)--represents an equivalent thickness of paint film
lost during the early stages of aging, ym
B = Constant (slope)--film erosion rate, ym/yr
Thus, if pollutants affect paint films, then different exposure conditions
should produce significantly different erosion rates. This hypothesis was
examined for each paint.
Oil Base House Paint
Visual examination of these panels revealed that all exposure conditions
caused considerable damage to this paint. In some cases the aluminum substrate
could be seen through the thinned paint film. Calculated erosion rates were
high and confirmed this observation. Table 28 presents the erosion rate values.
-------
Erosion rates ranged from a low of 28.3 to a high 79.1 ym/yr; the average was
60.0 ym/yr. Best estimate of the standard deviation on individual rates was
^7.1 ym/yr.
Table 29 gives the results of exposure to clean air at the tworcontrolled
relative humidities. Eroison rates ranged from 4.2 to 48.6 ym/yr and averaged
Table 28. EROSION RATES OF OIL BASE HOUSE PAINT
EXPOSED TO DESIGNATED CONTROLLED POLLUTED
AIR ENVIRONMENTAL CONDITIONS
Erosion rate and standard deviation, ym/yr
High relative humidity
High S02
65.4 + 7.0
79.1 + 5.4
72.8 + 7.2
61.4 + 3.4
Low S02
43.7 + 3.7
30.7* + 7.8
72.7 + 12.7
52.1 +_ 8.2
Low relative. humidity
High S02
45.8 + 1.9
52.6 + 3.8
52.4 + 3.9
60.6 + 2.8
Low S02
29. 4 +_2. 2
28.3 + 3.3
48.4 + 5.8
36.7 + 8.0
Exposure
condition
High 03
Low 03
High 03
5
Low 03
High N02
Low NO 2
Note: Erosion rates based on nine data sets per exposure condition except where
noted; *six data sets
Table 29. EROSION RATES OF OIL BASE HOUSE'PAINT EXPOSED
TO DESIGNATED CONTROLLED CLEAN AIR ENVIRONMENTAL CONDITIONS
Erosion rate and standard
deviation, ym/yr
High relative
humidity
48.6 + 9.5
4.2 + 2.7
Low relative
humidity
12.5 +_0.6
7.5 + 1.5
Exposure
condition
High
temperature
Low
temperature
Note: Erosion rates based on nine data sets
per exposure condition
50
-------
18.2 ym/yr. Best estimate of the standard deviation on individual rates was
+5.0 ym/yr. Both average erosion rates and standard deviations were lower
when exposed in clean air than in polluted air.
Analysis of variance of the erosion rates are given in Table 30. With
little risk, one may conclude that S02 concentrations and relative humidity
significantly affected the rates of erosion of oil base house paint. The
presence of N02 apparently increased the weight of the paint film; the amount
decreased as the S02 level increased. These two effects (direct N02 effect
and S02 x N02 interaction effect) were not as significant as the S02 and
relative humidity effects. These four effects accounted for nearly 81 percent
of the between-chamber variability; relative humidity and sulfur dioxide alone
accounted for 61 percent.
Table 30. ANALYSIS OF VARIANCE OF OIL BASE HOUSE PAINT
EROSION RATES FOR POLLUTED AIR EXPOSURE CONDITIONS
Factor
S0?
i.
RH
°3
N02
S02 x RH
S02 x 0,
S02 x N02
RH x 03
RH x N02
0., x NOp
Residual
TotaJ
Contrast
9.272
7.768
1.857
-5.097
0.661
-4.018
4.556
2.061
0.079
-2.644
Sum of
squares
1375.483
965.500
55.168
415.650
6.983
258.325
332.060
67.939
0.101
111.884
240.666
3830.059
Degrees
of freedom
1
1
1
1
1
1
1
1
1
1
5
15
Mean
squares
1375.483
965.600
55.168
415.650
6.983
258.325
332.060
67.939
0.101
111.884
48.133
F
28.58*
20.06**
1.15
8.64*
0.15
5.37
6.90*
1.41
0.00
2.32
R2
0.35
0.252 .
0.014
0.109
0.002
0.067
0.087
0.018
0.000
: 0.029
0.062
Notes: F uses the residual mean square in the denominator
R2 is the coefficient of determination excluding the within chamber variability.
** 99 percent probability level of significance
* 95 percent probability level of significance
51
-------
A multiple linear regression on S02 concentration and relative humidity
yielded the following relationship:
E = 14.3 + 0.0151 S02 + 0.388 RH. (10)
where E = erosion rate, yra/yr
S02 = concentration of S02, yg/m3
RH = relative humidity, %
The 95 percent confidence limits on the calculated erosion rates were ±
23.1 ym/yr. This empirical function predicts that under clean air conditions
at 90 percent and 50 percent relative humidity, oil base house paint erosion
rates should be 49.2 and 33.7 ym/yr, respectively, within"the confidence
limits of ± 23.1 ym/yr, 95 percent of the time. •
Results of the clean air exposures were consistent with this prediction.
A multiple linear regression of the clean air exposure data produced:
3
E = 37.2 + 0.405 RH + 1.133 T (11)
where E = erosion rate, ym/yr
RH = relative humidity, %
T = temperature, °C
The coefficient for the relative humidity effect agrees closely with the
similar value in the polluted air function. The clean air function predicts
that at 35°C and 90 percent and 50 percent relative humidities, .erosion
rates should be 38.9 and 22.7 ym/yr, respectively. Both values are well
within confidence limits of the polluted air function.
A simpler regression on the product of relative humidity and temperature,
which produced a better fit of the clean air data, yielded a linear relationship:
E = 11.8 + 0.0179 RH x T ' (12)
This function accounted for 89 percent of the variability as compared with
70'percent for the multiple linear regression. At 35°C and 90 percent and
50 percent relative humidities, calculated erosion rates were 44.7 and 19.6
ym/yr, respectively. These values are even more consistent with the predicted
values from the polluted air exposure conditions. The interaction effect
of input relative humidity and temperature is consistent with the fact that
both must be set to establish a level of moisture in the air. When the
moisture level (absolute humidity) remains constant, the relative humidity
must decrease if the temperature increases.
52
-------
Acrylic Latex House Paint
Blisters appeared on the panels exposed to the high S02 levels for
the first eight exposure conditions. These blisters were examined under a
scanning electron microscope and a light microscope. They were caused by
a build up of aluminum corrosion products at pits in the substrate. Microprobe
analyses revealed high concentrations of sulfur in the corrosion products.
The crystalline structure of the material and its water solubility indicated
that the product was aluminum sulfate rather than sulfate contaminated aluminum
oxide.
Acrylic latex house paint is primarily formulated for application on
non-metallic substrates; however, it can be applied to primed metallic surfaces.
A desirable characteristic of this paint is permeability to water vapor; this
allows a house to "breathe" when considerable differences exist between inside
and outside temperatures and relative humidities. Paints that do not have
this characteristic may peel because of a build up of condensed moisture between
the substrate and film. For this exposure study unprimed aluminum was selected
as the substrate because (1) variations in moisture absorption by wood would
mask any changes in weight of the paint film, (2) the thickness tester requires
an electrical conducting substrate, (3) aluminum is relatively corrosion resistant
when boldly exposed to polluted environments, and (4) aluminum is an excellent
thermal conductor (to accelerate the formation of dew).
The severe pitting of the aluminum under the paint film was unexpected.
The coating may have served as a semipermeable membrane retaining moisture
under the surface and excluding oxygen that would tend to passivate the alum-
inum.
Because of the attack on the substrate, exposures of the acrylic latex
house paint in this study were terminated after the completion of the first
eight exposure conditions. Stainless steel will be used with this paint in
future experiments. Pitting of the substrate beneath the paint film produced
significant weight increases that are not a measure of film erosion rate.
Therefore, these data were invalid and are not presented.
Vinyl Coil Coating
Visual appearance of this factory applied paint film indicated no damage.
Calculated erosion rates for the polluted exposures (Table 31), however, showed
that the film eroded but at rates considerably less than the oil base house
paint. Values ranged from a low of 1.43 to a high of 5.34 ym/yr; the average
erosion rate was 3.29 ym/yr. Best estimate of the standard deviation on individual
erosion rates was ±0.49 ym/yr.
Calculated erosion rates for the clean air exposures are given in Table
32. Erosion rates ranged from 0.114 to 2.49 ym/yr and averaged 1.29 ym/yr.
53
-------
Table 31. EROSION RATES OF VINYL COIL COATING EXPOSED TO
DESIGNATED CONTROLLED POLLUTED AIR ENVIRONMENTAL
CONDITIONS
Erosion rate and standard deviation, pm/yr
High relative humidity
High S02
4.76 ± 0.44
4.52 ± 0.62
5.34 ± 0.44
4.25 ± 0.71
Low S02
4.67 ± 0.46
1.43 ± 0.57
2.43 ± 0.22
1.43 ± 0.36
Low relative humidity
High S02
1.85 ± 0.4.1
2.33 ± 0.25
3.15 t 0.25
4.53 ± 0.42
Low S02
2.12 ± 0.34
2.11 ± 0.24
3.67 ± 0.68
4.02 ± 0.70
Exposure
condition
High 03
Low 0,
High 03
Low 03
High N02
Low N02
Note: Erosion rates based on nine data sets per exposure condition.
Table 32. EROSION RATES OF VINYL COIL COATING EXPOSED TO
DESIGNATED CONTROLLED CLEAN AIR ENVIRONMENTAL
CONDITIONS
Erosion rate and
standard deviation, ym/yr
High relative
humidity
2.01 ± 0.27
0.114 ± 0.088
Low relative
humidity
2.49 t 0.45
0.558 ± 0.218
Exposure
condition
High
temperature
Low
temperature
Note: Erosion rates based on nine date sets per exposure condition.
54
-------
Best estimate of the standard deviation on individual erosion rates was ±
0.287 ym/yr. On the basis of the calculated average erosion rate, vinyl coil
coatings would have a life expectency of 20 years in clean air environments.
Results of the analysis of variance are given in Table 33. Four factors
appeared to increase the erosion rate of the vinyl coil coating: (1) S02
concentrations, and relative humidity interacting individually with concentra-
tions of (2) S02, (3) 03, and (4) N02. These four factors accounted for 70.6
percent of the non-error (within chamber) variability.
Simple linear regression on the product of relative humidity and S02
concentration yielded
E = 2.51 + 1.60 x 10"5 x RH x S02
where E = erosion rate, ym/yr
RH = relative humidity, %
S02 = concentration of S02, yg/m3
Table 33. ANALYSIS OF VARIANCE OF VINYL GOIL COATING EROSION
RATES FOR POLLUTED EXPOSURE CONDITIONS
(13)
Factor
so2
RH
°3
S02 x Rd
S02 x 03
S02 x N02
RH x 03
RH x N02
03 x N02
Residual
Total
Contrast
0.5543
0.3159
0.2103
-0.3136
0.5602
-0.2762
-0.1626
0.4859
0.5556
0.1664
Sum of
squares
4.916
1.597
0.708
1.573
5.021
1.220
0.423
3.778
4.938
0.443
1.745
26.362
Degrees
of freedom
1
1
1
1
1
1
1
1
1
1
5
15
Mean
squares
4.916
1.597
0.708
1.573
5.021
1.220
0.423'
3.778
4.938
0.443
0.349
F
14.09*
4.58
2.03
4.51
14.39*
3.50
1.21
10.83*
14.15*
T.27.
R2
0.186
0.061
0.027
0.060
0.190
0.046
0.016
0.143
0,187
0.017
.0.066
Notes: F uses the residual mean square in the denominator
O
R is the coefficient of determination excluding the within chamber visibility
* 99 percent probability level of significance
55
-------
This empirical function accounted for 34 percent of the variability and RH x
S02 was the only factor that was statistically significant at the 95 percent
probability level. The function indicates that at the primary air quality
standard of 80 yg/m3 for S02, this pollutant only contributes between 2.5 and
4.4 percent to the erosion of vinyl coil coatings. Furthermore, the polluted
air intercept (2.51) is consistent with'the average erosion rate (2.25) for
the clean air exposures conducted at the same input temperature of 35°C.
The differences between the two values is not statistically significant.
Acrylic Coil Coating
Like the vinyl coil coating, the acrylic coil coating retained its original
appearance during exposures. Calculated erosion rates (Table 34) for the
polluted exposure conditions were low and ranged from 0.74 to 1.27 ym/yr.
The average erosion rate was only 0.57 ym/yr. Best estimate of the standard
deviation on individual erosion rates was ±0.31 ym/yr. Even under the most
severe exposure conditions, this particular film can be expected to last
at least 15 years.
Table 34. EROSION RATES OF ACRYLIC COIL COATING EXPOSED TO
DESIGNATED POLLUTED AIR ENVIRONMENTAL CONDITIONS
Erosion rate and standard deviation, ym/yr
High relative humidity
High S02
0.83 ± 0.26
0.55 ± 0.32
1.25 ± 0.71
0.25 •:' 0.08
Low S02
1.27 ± 0.20
0.43 ± 0.25
1.15 ± 0.15
0.74 ± 0.24
Low relative humidity
High S02
0.42 ± 0.13
0.66 ± 0.25
0.41 ± 0.39
0.52 ± 0.28
Low SOp
1.29 ± 0.37
-0.11 ± 0.15
0.26 ± 0.40
0.60 ± 0.14
Exposure
condition
High 03
Low 03
High 03.
Low 03
High N02
Low N02
Note: Erosion rates based on nine data sets per exposure condition.
56
-------
Table 35 gives the calculated erosion rates for the clean air exposure
conditions. Erosion rates ranged from 0.48 to 0.20 ym/yr and averaged 0.09
ym/yr. Best estimate of the standard deviation on individual erosion rates
was ± 0.62 ym/yr. On the basis of the calculated average erosion rate, acrylic
coil coatings should have a life expectency of over 200 years when exposed
in clean air.
Results of the analysis of variance for the polluted air exposures are
given in Table 36. Ozone was the most likely factor to affect erosion rates
of acrylic coil coatings. A linear regression of the erosion rate data as
a function of 03 level yielded
E = 0.159 + .000714 03
(14)
where E = erosion rate, ym/yr
03 = concentration of 03, yg/m3
The 95 percent confidence limits on the data were ± 0.98 ym/yr. Although
the effect of ozone was statistically significant, for all practical purposes
the effect was negligible.
Analysis of variance of the clean air exposure data indicated that neither
relative humidity nor temperature had a significant effect on the erosion
rate of acrylic coil coatings. The mean value of 0.09 ym/yr was consistent
with an erosion rate of 0.52 ym/yr predicted by the empirical function derived
from polluted air exposure date.
Table 35. EROSION RATES OF ACRYLIC COIL COATING EXPOSED TO
DESIGNATED CONTROLLED CLEAN AIR ENVIRONMENTAL CONDITIONS
Erosion rate and
standard deviation, ym/yr
High relative
humidity
0.13 ±0.15
0.08 ± 0.12
Low relative
humidity
0.20 ± 1.20
-0.05 ± 0.19
Exposure
condition
High
temperature
Low
temperature
Note: Erosion rates based on nine data sets per exposure condition.
57
-------
Table 36. ANALYSIS OF VARIANCE OF ACRYLIC COIL COATING
EROSION RATES FOR POLLUTED EXPOSURE CONDITIONS
Factor
so2
RH
°3
N02
S02 x RH
S09 x 0,
t 0
S02 x N02
RH x 03
RH x N02
0, x NO,
•J £.
Residual
- Total
Contrast
0.0456
0.0587
0.2941
0.1028
0.499
-0.1784
-0.0989
0.2049
0.0424
-0.0106
Sum of
squares
0.0332
0.0551
1.3836
0.1691
0.0399
0.5094
0.1566 •
0.6720
0.0288
0.0018
1.2582
4.3077
Degrees
of freedom
1
1
1.
1
1
1
1
1
1
1
51
15
Mean
squares
' 0.0332
0.0551
1.3836
0.1691
0.0399
0.5094
0.1566
0.6720
0.0288 '
0.0018
0.2516
F
0.13
0.22
5.50
0.67
0.16
2.02
0.62
2.67
0.11 '
0.01
R2-
0.008 '
0.013 -
0.321
0.039 '
0.009
0.118
0.036
0.156
0.007
0. 000
'0.292
Notes: F uses the residual mean square in the denominator
2
R 1s the coefficient of determination excluding the within chamber variability
58
-------
SECTION XII
EFFECTS ON DYED FABRICS
Drapery fabrics were selected for exposure because they represent textiles
that are both economically important and subject to fading. Furthermore,
they are designed to have fairly long life; however, their useful life may
be limited by accumulated fading. Draperies are normally exposed indoors
to daily cycles of sunlight and to pollution levels that are generally lower
than those existing outdoors.
METHODOLOGY
A large producer of standard-size draperies, which are distributed and
sold nationwide, supplied the study with samples of three fabrics (Table
37). The fabrics were popular colors and represented moderate to high volume
usage.
Table 37. DESCRIPTION OF DRAPERY FABRICS
Fabric Color
66% Rayon, 34% Acetate Royal Blue
66% Rayon, 34% Acetate Red
100% Cotton duck Plum
The identity of the dyes on each fabric was unknown. A superficial
evaluation, however, revealed that none of the fabrics contained the vulnerable
disperse dyes known to fade in the presence of atmospheric nitrogen oxides
and/or ozone. The plum colored cotton duck fabric, which faded considerably
as later exposure results will show, appeared to be vat dyed.
The drapery fabrics were cut into 8.89 x 6.35 cm panels, mounted in
plastic frames, and conditioned for 48 hours at constant temperature (25
± 0.5°C) and relative humidity (45 ± 1%). The initial color of each fabric
panel was then analyzed photoelectrically with a Hunter Model D25A color
difference meter. The instrument measures color as a three dimensional point
(L,a,b) in space. The (L) value is a measure of lightness, the (a) value
of red to green, and the (b) value yellow to blue. Fading is expressed in
color difference or fading units (AE) and was calculated using the equation:
59
-------
AE = /(Lo-L)2 + (ao-a)2 + (bQ-b)2 (15)
where the subscript "o" refers to initial values.
Beloin describes the color analyzing technique in more detail.16
Three replicate panels of each fabric were randomly placed in the exposure
chambers in such a way that moisture would not condense on them during the
dew cycle. (Moisture is unlikely to condense on household textiles during
normal use.) After exposure periods of 250, 500 and 1000 hours, the panels
were removed and conditioned at constant temperature and relative humidity,
analyzed for color, and, except for the final exposure period, returned to
the chambers for additional exposure.
RESULTS AND DISCUSSION
After 1000 hours exposure, the most severe condition--high pollution
levels and high relative humidity--caused maximum fading of the drapery fabrics.
Color change (AE) values for the royal blue, red, and plum colored fabrics
were 2.4, 8.7, and 29.1 respectively. A color change less than 3.0 units
is difficult to detect by eyesight.
Because most of the fading data appeared to be linearly proportional
to exposure time over these relatively short exposure periods, least squares
fits of the fading data versus exposure time (through the origin) were calculated.
The resulting slopes represent fading rates. Table 38 presents the fading
rates for the three drapery fabrics and 16 exposure conditions. Fading rates
ranged from a low of 16 fading units per year for the royal blue fabric,
to a high of 301 fading units per year for the plum fabric. These values
represent initial fading rates and not the amount of fading that would be
expected to occur in a year. Under actual long term exposures, fading is
a non-linear function of time and fading rates approach zero with time.
The fabrics were also exposed to clean air under different input relative
humidity and temperature conditions. These results are given in Table 39;
fading rates ranged from a low of 11 to a high of 94 units per year.
Results of an analysis of variance for the three fabrics exposed to
the polluted air environments are given in Tables 40, 41, and 42. Using
all triple and greater interactions to calculate the residual or error mean
square, only relative humidity was a significant factor for both the red
and royal blue fabrics. The significant factors for the plum fabric were
relative humidity, N02, and the interation between relative humidity and
NC>2. For all three fabrics, changing (increasing) the input relative humidity
produced the greatest effect.
60
-------
Table 38. INITIAL FADING RATES OF DRAPERY FABRICS
EXPOSED TO DESIGNATED CONTROLLED POLLUTED
AIR ENVIRONMENTAL CONDITIONS
Drapery
fabric
color
Blue
Red
Plum
Blue
Red
Plum
Blue
Red
Plum
Blue
Red
Plum
Initial fading rate and standard
deviation, fading units per year
High relative humidity
High S02
21 ± 10
73 ± 9
301 ± 89
28 ± 15
80 ± 12
261 ± 38
26 ± 8
47 ± 7
135 ± 19
35 ± 14
98 ± 10
142 ± 20
Low S02
27 ± 10
103 ± 8
201 ± 6
24 ± 13
96 ± 11
203 ± 7
27 ± 11
102 ± .7
146 ± 13
37 ± 15
91 ± 5
131 ± 13
Low relative humidity
High S02
18 ± 7
60 ± 3
120 ± 20
20 ± 5
63 ± 7
106 ± 4
23 ± 6
66 ± 5
90 ± 5
17 ± 6
51 ± 4
68 ± 5
Low S02
17 ± 3
83 ± 12
120 ± 4
19 ± 1
81 ± 3
122 ± 13
16 ± 3
38 ± 6
94 ± 5
23 ± 8
33 ± 3
80 ± 6*
i
Exposure condition
High 03
Low 0-
High 03
Low 0-
High N02
Low N02
Note: Fading rates based on nine data sets per exposure condition
except where noted; * 8 data sets.
61
-------
Table 39. INITIAL FADING RATES OF DRAPERY FABRICS
EXPOSED TO DESIGNATED CONTROLLED CLEAN
AIR ENVIRONMENTAL CONDITIONS
Drapery
fabric
color
Blue
Red
Plum
Initial fading rate and standard
deviation, fading units per year
High relative humidity
High
temperature
18 ± 10
75 ± 13
94 ± 12
Low
temperature
12 ± 7
16 ± 3
68 ± 8
Low relative humidity
High
temperature
15 ± 5
26 ± 2
48 ± 5
Low
temperature
11 ± 4
14 ± 6
60 ± 7
Note: Fading rate based on nine data sets per
exposure condition.
62
-------
Table 40. ANALYSIS OF VARIANCE OF ROYAL BLUE FABRIC FADING
RATES FOR POLLUTED AIR EXPOSURE CONDITIONS
Factor
so2
RH
°3
N02
S02 x RH
SO, x 0,
k O
S02 x N02
RH x 03
RH x N02
03 x N02
Residual
Total
Contrast
-0.0106
4.5769
-1.7506
-1.8731
-0.5094
0.2406
0.1256
-1.1119
-1.2019
0.9006
Sum of
squares
0.0018
335.1646
49.0350
56.1376
4.1514
0.9264
0.4935
19.7806
23.1121
12.9780
78.6483
580.4293
Degrees
of freedom
1
1
1
1
1
1
1
1
1
1
5
15
Mean
square
0.0018
335.1646
49.0350
56.1376
4.1514
0.9264
0.4935
19.7806
23.1121
12.9780
15.7297
F
0.00
21.31**
3.12
3.57
0.26
0.06 *
0.003
1.26
1.47
0.83
R2
0.000
0.577
0.085
0.097
0.007
0.002
0.001
0.034
' 0.040
0.022
0.136
** 99 percent probability level of significance
63
-------
Table 41. ANALYSIS OF VARIANCE OF RED FABRIC FADING RATES
FOR POLLUTED AIR EXPOSURE CONDITIONS
Factor
so2
RH
°3
N02
S02 x RH
S02 x 03
S02 x N02
RH x 03
RH x N02
°3 * N02
Residual
Total
Contrast
-5.505
13.471
-1.265
7.020
-6.320
-4.456
-5.291
-3.555
-5.429
1.141
Sum of
squares
484.880
2903.593
25.604
788.486
639.078
317.731
447.957
202.208
471.324
20.839
1452.827
7754.527
Degrees
of freedom
1
1
1
1
1
1
1
1
1
1
5
15
Mean
square
484.880
2903.593
25.604
788.486
639.078
317.731
447.957
202.208
471.324
20.839
290.5654
F
1.67
9.99*
0.09
2.71
2.20
1.09 *
1.54
0.70
1.62
0.07
R2
0.063 '
0.374
0.003
0.102
0.082
0.041
0.058
0.026
\
' 0.061
0.003
0.187
* 95 percent probability level of significance
64
-------
Table 42. ANALYSIS OF VARIANCE OF PLUM FABRIC FADING
RATES FOR POLLUTED AIR EXPOSURE CONDITIONS
Factor
so2
RH
°3
N02
S02 x RH
S02 x 03
S02 x N0£
RH x 03
RH x N02
°3 x N62
Residual
Total
Contrast
8.065
44.925
5.768
34.193
11.840
2.768
9.848
-0.085
17.200
0.373
Sum of
squares
1040.708
32292.091
532.225
18706.033
2242.970
122.545
1551.572
0.116
4733.440
2.220
2255.571
63479.491
Degrees
of freedorr
1
1
1
1
1
1
1
1
1
1
5.
15
Mean
square
. 1040.708
32292.091
532.225
18706.033
2242.970
122.545
1551.572
0.116
4733.440
2.220
451.114
F
2.31
71.58**
1.18
41.47**
4.97
0.27
3.43
0.00
10.49*
0.00
R2
0.016
0.509
0.008
0.295
0.035
0.002
0.024
, 0.00
0.075
0.000
0.036
* 95 percent probability level of significance
* 99 percent probability level of significance
65
-------
Results of an analysis of variance of the clean air exposure fading
data are given in Table 43. When the within-sample-error to calculate values
of F was used, the direct effects and the interaction between temperature
and relative humidity were all significant at the 99 percent probability
level. Using the more stringent test of. assuming the triple interaction
as the residual error, only the obvious color differences among the fabrics
were significant.
The plum fabric, therefore, was the only drapery fabric significantly
affected by any of the pollutants; nitrogen dioxide contributed direct.ly
and synergistically with relative humidity to increased fading rates. The
amount of fading that occurred after 1000 hours of exposure was easily detected
visually. Thus, pollutant fading of this drapery fabric could have economic
significance if the magnitude of fading under expected ambient conditions
is sufficiently large. A dose-response relationship is needed to calculate
this magnitude.
The rate of fading at any point in time is expected to be proportional
to the concentrations of the reacting ingredients. These ingredients are
energy (light and heat), moisture, N0£, and dye. During the chamber exposures,
all ingredient levels except the dye remained constant with time. At infinite
time all the dye that is available to react will have reacted to produce
maximum fading (AEm). Thus, AEm is a measure of the total amount of dye
available to react before fading, and (AEp - AE) is a measure of the amount
available at any time (t). For any particular set of conditions:
Table 43. ANALYSIS OF VARIANCE OF FADING RATES
FOR CLEAN AIR EXPOSURE CONDITIONS
Factor
RH
T
Fabric
RH x T
Sum of
squares
RH x Fabric
T x Fabric
RH
Wi
**
*
x T x Fabric
thin Sample
99 percent
95 percent
Error
probabil
probabil
987.
751.
5966.
637.
399.
588.
290.
5233.
ity 1
ity 1
361
292
859
000
897
995
201
250
eve!
eve!
Degrees
of freedom
1
1
Mean
square
987
751
1 2983
1
2
2
2
96
of significance
of significance
637
199
294
145
54
.361
.292
.430
.000
.449
.491
.101
.513
18
13
54
11
3
5
2
F
.11**
.78**
.73**
.69**
.67
.40*
.66
0
0
0
0
0
0
0
0
R2
.0665
.0506
.4017
.0429
.0269
.0397
.0195
.3523
66
-------
= a0 (AE - AE) (16)
m
where —rr = the fading rate.
a0 = a constant that is a function
of light, moisture in the air,
and NC>2 level.
The amount of fading as a function of time, therefore, is:
AE = AEm (l-e'*'*) (17)
The actual fading data for the plum fabric indicate that the following
empirical function is a good approximation:
AE = 30[l-e-C°-75 + °'01M * 2'9 x 10"x N°2 ] (18)
where AE = amount of fading, fading units
M = amount of moisture, mg/m3, at 25°C and one atmosphere
(determined by input relative humidity and temperature)
NC»2 = concentration of N02, yg/m3
t = exposure time, yr.
Despite the total number of factors that were varied and the simplicity
of the approximation, this function accounts for nearly 78 percent of the
variability. The degree of fit is shown in Figure 11.
Unacceptable levels of fading are subjective and vary considerably with
individuals. Nevertheless, the presence of N02 and moisture in the ambient
air will synergistically accelerate fading and thus reduce the time required
to reach any unacceptable level. This effect can be conveniently expressed
in terms of percentage useful life lost. By letting tj and t2 represent the
times necessary to reach unacceptable fading in ambient clean air and in ambient
air containing N02, respectively, then
Ai - t2\
Percentage Life Lost = I 100 (19)
Since the level of unacceptable fading is the same for both exposure time
periods,
67
-------
30
24
t/l
4J
•r—
C
cn
c
•o
12
UJ
<
O
<:
a
UJ
oi
\ I
12
24
30
PREDICTED FADING VALUE, &E fading units
Figure 11. Comparison of predicted and measured fading
values for the plum colored fabric exposed to laboratory
controlled polluted air and clean air conditions.
68
-------
AE = E.
m
(•
1-e
= E
m
(-<
1-e
(20)
(21)
Therefore,
100 =
(22)
Substituting for a0 the expression 0.75 + 0.01M + 2.9 x 10'5 x N02 x M and
remembering that time tj corresponds to exposure in clean air containing no
N02, the following relationship is derived:
Percentage Life Lost =
2.9 x IP"3 x N02 x M
0.75 + 0.01M + 2.9 x 10-5 x N02 x M
(23)
When the plum fabric is exposed to the amount of light used in these
exposures, the value 0.75 is a factor associated with fading by light. Since
draperies are seldom exposed indoors to such intense light (simulating midday
direct sunlight), arbitrarily reducing the factor by 1/2 should provide a
more realistic relationship. At 60 percent relative humidity and 25°C, the
amount of moisture in the air is approximately 115 mg/m3. Equation (23), there-
fore, becomes
0.45 N02
(24)
Percentage Life Lost = 1>93 /0>0045
Figure 12 is a plot of this function. It predicts that at a concentration
of N02 equivalent to the national primary ambient air quality standard of
100 yg/nr would reduce the useful life of the plum fabric by 19 percent.
Because only one of three fabrics was appreciably faded by any of the
three pollutants, product selection rather than pollution abatement is probably
the most economic solution to the fading problem. Before consumers can make
this selection, however, textile producers must make them aware of the fading
characteristics of fabrics by proper labelling.
69
-------
J_
O)
a.
GO
o
80
70
60
50
40
30
20
10
50
100
150
200
250
NITROGEN DIOXIDE CONCENTRATION,
Figure 12. Effect of nitrogen oxide concentration
on the fading of the plum colored fabric.
70
-------
SECTION XIII
EFFECTS ON ELASTOMERS AND PLASTICS
Elastomeric and plastic products are economically important materials
that air pollutants could potentially damage and cause a loss in useful life.
Two representative and economically significant products were selected for
exposure in this study: white sidewall rubber from radial ply tires and extruded
white vinyl house siding.
Scientists and engineers have known for many years that atmospheric ozone
deteriorates products made from a number of common elastomers. Small cracks
that progressively enlarge develop in areas under stress; this effect is called
"ozone cracking." The addition of antiozonants to elastomeric formulations
offers limited protection against ozone cracking. For rubber tires, this
protection is usually sufficient for the useful life of the product.
Because of staining, however, antiozonants are not added to white sidewall
rubber. Tire manufacturers normally depend on the use of more resistant elastomers
for protection. Over the years this approach has been satisfactory for conventio-
nal non-radial tires, but may not hold true for radial tires since tread wear
is about twice as long. Therefore, white sidewall failure as a result of
ozone cracking could be a limiting factor in the life of radial tires.
Vinyl house siding is a frequently used product because of its long life
and essentially maintenance free surface. The presence of air pollutants,
however, may cause damage that would reduce the useful life of this product.
METHODOLOGY
Rubber
White sidewall rubber specimens were cut from a top-of-the-line steel
belted, rayon cord, radial ply tire. The size of each specimen strip was
1 cm x 9 cm, with the long dimension parallel to the axis of the cord and
perpendicular to the white wall strips. The cut edges of the specimens were
coated with wax to prevent absorption of moisture and pollutants by the exposed
ends of rayon cord.
Special aluminum jigs were used to hold three rubber specimens under
a predetermined strain during exposure. For each exposure condition, three
specimens were exposed at 10 percent strain and three at 20 percent strain.
71-
-------
The jigs were randomly mounted in the chambers so that dew did not condense
on the rubber specimens. Nevertheless, the specimens were exposed to changes
in temperature and relative humidity caused by the dew/light cycle.
After each nominal exposure period of 250, 500, and 1000 hours, photomacro-
graphs (lOx) were made of the specimens while still under stress. On completion
of the 1000-hour exposures, dumbbell tensile specimens were die cut from
the exposed strips. Tensile tests were then carried out using an Instron
tensile testing machine to determine the breaking strength of the rayon
cord.
Vinyl Siding
Specimens measuring 5 cm square were cut from white vinyl house siding
and randomly exposed in triplicate. Because appearance is a primary factor
limiting the useful life of this product, a change in reflectance was
used to assess potential surface effects. Reflectance was measured with
a photovolt meter using blue, green and amber filters.. For each of the 16
exposure conditions, measurements were taken initially and after each nominal
exposure period of 250, 500, and 1000 hours. Specimen surfaces were also
examined visually for possible effects.
RESULTS AND DISCUSSION
Rubber
As a result of die cutting the exposed rubber strips into dumbbell tensile
specimens, the number of uncut rayon cords per specimen varied from 9 to
12, thus affecting the breaking strength of the specimens. To reflect this
variation, tensile test responses were divided by the number of rayon cords
and reported as kilograms per cord.
Table 44 presents the tensile test results for the first eight exposure
conditions, which included low and high levels of S02, 03 and relative humidity,
and low levels of N02. Theoretically, pollutants should have no effect on
the cord unless ozone cracking penetrates to cord depth. Therefore, an analysis
of variance was made on these results as a group. If the analysis showed
that the factors --strain, S02, 03 and relative humidity--did not significantly
reduce the strength of the cord, tensile tests on specimen exposed to the
remaining conditions (high levels of N0£) could be omitted.
Although the results in Table 44 are incomplete because it includes
data from only seven of the eight exposure conditions> an analysis of variance
can be made for all but two possible interaction effects. To avoid experimental
72
-------
Table 44. CORD BREAKING STRENGTH OF WHITE SIDEWALL
RUBBER TIRE SPECIMENS EXPOSED UNDER STRESS
TO DESIGNATED CONTROLLED POLLUTED AIR
ENVIRONMENTAL CONDITIONS FOR 1000 HOURS
Strain
20%
10%
20%
10%
Cord breaking strength and
standard deviation, kg/cord
High relative humidity
High S02
8.3±0.8
8.3±0.3
8.2±0.7
11.1+2.1
Low S02
8.1+0.8
8.0±0.6
9.2±0.8
8.6±0.2
Low relative humidity
High S02
9.6±0.8
9.5+0.3
9.2±1.8
8.3±0.6
Low S02
8.2±0.8
8.9+0.3
Invalid
data
Invalid
data
Exposure
condition
High 03
Low 03
Notes: 1.
2.
Cord breaking strength values based on three data
sets per exposure condition.
Low N02 levels were constant for all eight
exposure conditions.
bias, however, error mean square was calculated using interaction-effects-
variance rather than within-sample-variance. The reasons for this were
(1) the three specimens exposed per chamber were grouped together in a single
aluminum jig and thus individual specimens could not be placed randomly,
and (2) each exposure condition was not replicated. Table 45 gives the analyses
of variance for the direct effects.
73
-------
Table 45. ANALYSIS OF VARIANCE OF TIRE CORD BREAKING
STRENGTH FOR STATISTICALLY SIGNIFICANT FACTORS
Factor
Relative
humidity
S02
Stress
°3
Error
Sum of
squares
0.2408
1.3333
0.2579
0.6533
6.8982
Degrees of
freedom
1
1
1
1
9
Mean
square
0.2408
1.3333
0.2579
0.6533
0.7664
F
0.314
1.740
0.337
0.852
None of these factors caused a statistically significant decrease in
cord breaking strength, most likely because the pollutants never made contact
with the cord. Microscopic examination of the pulled specimens revealed
that environmentally induced cracks in the white sidewall rubber did not
penetrate to the cord. Thus, rubber specimens exposed to the remaining eight
conditions were not subjected to tensile tests.
Figure 13 is a composite macrograph (about 10X) showing the cracking
effects that the white sidewall rubber tire specimens developed when exposed
under high stress to the 16 polluted air exposure conditions for 1000 hours.
Specimens showed a similar but less severe pattern when exposed under low
stress.
To assess the degree of rubber cracking, various quantitative methods
were considered. The number of cracks intersecting a given length, average
crack length within a given area, and summation of crack lengths within a
given area were rejected as valid measures of reduced product life. Tire
failure should occur at the weakest point. Thus, an extreme rather than
an average value should be more meaningful.
Maximum crack length for each macrograph of exposed specimen was used
as a measure of damage response. A least squares fit through the origin
of the crack length as a function of exposure time was calculated to get
a cracking rate. Table 46 shows the mean cracking rates for all 16 exposure
conditions and two strain levels. Values ranged from a low of 0 to a high
of 3.8 micrometers per hour (ym/hr).
An analysis of variance of these data revealed that only four of 31 possible
direct and synergistic effects were statistically significant at the 95 percent
confidence level. Included among the insignificant direct and synergistic
effects were the two levels of strain; normally, cracking is a direct function
of applied strain. Table 47 presents the analysis of variance for the four
statistically significant factors.
74
-------
HIGH N02
LOW N02
HIGH RELATIVE HUMIDITY
HIGH S02
LOW S02
LOW RELATIVE HUMIDITY
HIGH S02
LOW S02
HIGH 0
' :„•
LOW03
HIGH 03
i
^T^.~
•.XT /•
'• 'l'1 \ »
•'
LOW03
i
••\ (11
i,v
Figure 13. Macrographs of cracks developed by white sidewall rubber tire
specimens when exposed under high stress to designated controlled polluted
air environmental conditions for 1000 hours.
75
-------
Table 46. CRACKING RATES OF WHITE SIDEWALL RUBBER TIRE
SPECIMENS EXPOSED UNDER STRESS TO DESIGNATED
CONTROLLED POLLUTED AIR ENVIRONMENTAL CONDITIONS
Strain
10%
202
102
20%
10%
20%
10%
20%
Mean cracking rate and
standard deviation, ptn/hr
High relative humidity
High S02
1.775
±1.134
2.703
±0.755
0.853
±0.676
0.?83
±0.325
3.394
±0.865
2.414
±0.438
0.0
0.0 .
Low SQ2
0.761
±0.638 '
2.361
±0.460
1.377
±0.477
0.883
±0.363
3.000*
±0.000
3.760
±2.909
0.0
0.684
±0.408
Low relative humidity
High S02
1.028
±0.413
1.408
±0.271
2.401
±1.139
1.925
±0.576
1.756
±0.721
2.938
±1.193
1.332
±0.614
1,941
±0.939
Low S02
2.600**
±0.963
3.275*
±1;425
2.168*
±2.307
2.466
±0.576
3.768
±1.820
3.138
±0.679
1.331
±0.753
1 .809
±0.254
Exposure
condition
High 03
*
Low 03
High 03
Low 03
High N02
Low N02
Note: Cracking rate based on three data sets per exposure condition
.except where noted; * two data sets; ** one data set.
Table 47. ANALYSIS OF VARIANCE OF RUBBER CRACKING RATES
FOR STATISTICALLY SIGNIFICANT FACTORS
Factor
Relative
humidity (RH)
03
03 x RH
03 x N02
Error
Contrast
-0.32394
0.62269
0 . 33906
-0.44425
Degrees of
freedom
1
1
1
1
1 .
Mean
square
3.35794
12.40767
3.67883
6.31546
0.75403
F
4.45
16.46*
4.88
8.38*
*99 percent probability level of significance
76
-------
The rubber specimens did not show a loss in breaking strength because
the cracks, which developed during the relatively short exposure periods
(1000 hours) used in this study, were not deep enough to expose the cord.
As a result air pollutants could not make contact with the cord and cause
possible damage. Nevertheless, a loss in strength might have occurred if
cracking had been allowed to progress indefinitely.
As would be expected, 63 concentration was the major factor that accelerated
rubber cracking rates. In the absence of 03, increasing relative humidity
appeared to inhibit cracking; however, with 03 present increasing relative
humidity appeared to accelerate cracking. The presence of N0£ seemed to
reduce the damaging effects of 03.
In the real world, high levels of 63 and NC>2 generally occur together.
Also, the low levels used in this study represented the one hour national
primary ambient air quality standard for oxidant (03) and the annual average
primary ambient air quality standard for N02- The average rubber cracking
rate at these NC>2 and 63 levels was 0.89 ym/hr. Crack geometry was such
that the depth was less than one fourth the length. Thus, the depth cracking
rate into the sidewall would be less that 0.22 ym/hr or less than 2000 ym/yr.
For this particular tire, the depth of the white rubber was 3175 ym and
underneath was 1900 ym of black rubber protecting the cord. A developing
crack, therefore, must penetrate a distance of 5075 ym before exposing the
tire cord. Since black rubber contains antiozonants, the rate of cracking
may be less than for the white rubber. However, assuming the rate of cracking
for the black rubber to be the same as for the white rubber, it would take
about 2 1/2 years for cracks to penetrate to cord depth when continuously
exposed to air containing 03 and N02 at levels equivalent to national primary
ambient air quality standards. Additional time would be necessary for pollutants
to attack the cord.
The calculated crack penetration rate for the white sidewall rubber
was based on data derived from continuous exposures at primary ambient air
quality standard levels for ozone. But, since average annual 03 levels in
actual real world exposures are much lower, sidewall cracking for this particular
tire appears to be an unlikely cause of reduced tire life. Tire tread should
wear out first.
Other proprietary tires may be lower quality and produced with white
sidewall rubber that is less resistant to ozone. In this case sidewall cracking
could pose a problem. Lower quality tires, however, imply less tread wear
life. Consequently, tread wear, rather than sidewall failure, should also
determine the life of lower quality tires. Thus, the radial tire evaluated
in this study should be reasonably representative of all tires with respect
to air pollution effects.
77
-------
Vinyl Siding
Reflectance measurements on vinyl siding specimens essentially did not
change during any of the exposure conditions and thus revealed no detectable
damage response. Examination with a light microscope and a scanning electron
microscope of selected specimens exposed to the most severe environmental
conditions also failed to detect any surface damage.
This particular white vinyl house siding was, therefore, highly resistant
to damage by gaseous air pollutants as well as moisture and simulated sunlight.
Thus, other than soiling by airborne particles, common gaseous air pollutants
at ambient concentrations should have no effect on this product.
78
-------
SECTION XIV
EFFECTS ON MARBLE AND CEMENT
Marble and cement are widely used building materials. Marble is a natural
stone and one of the oldest structural as well as artistic materials used by
man. Cement is a universal manmade material used as a bonding agent for other
materials.
Down through the ages marble has shown a high degree of durability to
natural environments. Increasing world wide industrialization and use of
fossil fuels, however, have created air pollution. Pollutants not only obscure
the asethetic beauty of marble structures with soot and dirt, but also accelerate
surface erosion. Furthermore, researchers have found that sulfur dioxide environ-
ments promote the formation of water soluble salts which are easily leached away
by rain.17,18
Evidence that air pollutants attack cement has not been documented but is
suspected, mainly because cement contains substances that can react with acid
forming pollutants.
METHODOLOGY
Marble
White Cherokee marble with a hone finish was selected for exposure.
Specimens measured 2.54 x 2.54 x 0.67 cm.
Two techniques were used to evaluate the effects of pollutants: scanning
electron microscopy (SEM) with microprobe analysis, and loss-in-weight. SEM
analysis was used to assess evidence of possible surface reactions and sub-
surface reactions resulting from gaseous diffusion of pollutants (mainly S02).
To determine if diffusion occurred, x-ray microprobe line scans for sulfur
(concentration gradients) were made on cross sections of marble specimens.
Sulfur concentration gradients should increase with severity of exposure
condition.
Loss-in-weight was the difference between specimen weight before and after
1000 hours exposure. Weights were measured using an analytical balance accurate
to the nearest 0.1 mg.
Prior to weighing and exposure, specimens were cleaned and conditioned
for 48 hours at constant temperature (25°C) and relative humidity (45%). The
cleaning procedure consisted of washing in a 5 percent detergent solution
followed by rinsing in distilled water.
79
-------
For each exposure condition, three specimens were randomly placed on a
chamber exposure rack. At the conclusion of 1000 hours exposure, the specimens
were removed. Two of the specimens were cleaned, conditioned, and weighed;
the remaining specimen was examined by SEM techniques.
Cement
Cement specimens were cast from a water slurry of Portland cement in the
form of circular wafers about 2.5 cm in diameter and 0.3 cm thick. Loss-in-
weight was used to assess effects using the same methodology as described for
marble.
RESULTS AND DISCUSSION
Marble
Examination by SEM of marble specimens exposed for 1000 hours to the most
severe conditions (high pollutant levels) revealed that a surface reaction was
the primary damage mechanism. A photomicrograph (figure 14) of an exposed
surface showed crystalline material containing sulfur atoms, which were identified
by x-ray microprobe analysis (figure 15). Although qualitatative analysis
was not carried out, the crystalline material was probably calcium sulfate.
X-ray microprobe line scans for sulfur on cross sections of marble specimens
revealed penetrations to about 50 ym. However, diffusivity coefficients
calculated from sulfur concentration gradients suggested that the diffusion
mechanism did not significantly damage this particular marble.
Because SEM analysis was not quantitative, loss-in-weight was a better
technique for assessing the magnitude of damage caused by the pollutants. It
was assumed that water soluble reaction products removed from the surface of
exposed specimens during cleaning accounted for the loss-in-weight. Each
loss-in-weight value was converted to an equivalent thickness of calcium
carbonate and expressed as an erosion rate-loss in thickness units per year
(ym/yr). Table 48 presents average erosion rates (based on two data sets per
exposure condition) for all 16 exposure conditions.
An analysis of variance was made on the erosion rate data and the results
are shown in Table 49. Sulfur dioxide, relative humidity and ozone were
significant factors at the 95 percent probability level. These three factors
accounted for 65.7 percent of the variability.
80
-------
Figure 14. SEM photomicrograph (500X) of marble specimen
exposed for 1000 hours to high levels of pollutants and
high relative humidity.
Figure 15. Sulfur x-ray microprobe scan (500X) of marble
specimen shown in Figure 14.
81
-------
Table 48. EROSION RATES OF MARBLE EXPOSED TO DESIGNATED
CONTROLLED POLLUTED AIR ENVIRONMENTAL CONDITIONS
Erosion rate and standard deviation, ym/yr
High relative humidity
High S02
21.8 +. 9.1
9.1 + 10.4
8.6 +• 10.0
2.9 +_ 10.4
Low S02
7.4 ^4.9
5.7 +5.1
5.6 +6.0
3.5 + 3.4
Low relative humidity
High S02
4.2 +9.0
1.1 +4.6
7.6 + 12.9
3.4.+ 2.9
Low S02
5.0 +_2.5
4.1 +3.3
2.9 +2.7
3.8 +4.0
Exposure
Condition
High 03
A
Low 03
High 03
Low 0,
High N02
Low N02
Note: Erosion rate values based on two data sets per exposure condition.
Table 49. ANALYSIS OF VARIANCE OF MARBLE EROSION RATES
FOR POLLUTED AIR EXPOSURE CONDITIONS
Factor
so2
RH
°3
N02
SO, x RH
f.
S02 x 03
S02 x N02
RH x 03
RH x N02
03 x N02
Residual
Total
Contrast
1.814
1.561
1.325
0.526
0.728
1.066
0.143
0.093
0.914
-0.042
Sum of
squares
105.306
77.969
56.154
8.852
16.980
36.359
0.658 •
0.276
26.736
0.057
34.730
364.077
Degrees
of freedom
1
1
1
1
1
1
1
1
1
1
5
15
Mean
square
105.30.6
77.969
56.154
8.852
16.980
36.359
0.658
0.276
26.736
0.057
6.946
F
15.16*
11.23*
8.08*
1.27
2.44.
5,23
0.09
0.04
3.85 •-
0.01
R2
. 0.289
0.214
0.154
0.024
0.047
0.100
0.002
0.001
0.073
0.000
v 0.095
Notes: F uses the residual mean square in the denominator
2
•R is the coefficient of determination excluding the within chamber variability
* 95 percent probability level of significance
82
-------
Linear regression of these three factors produced the following relation-
ship:
E = -3.31 + 0.078 RH + 2.95 x 10"3 S02 + 3.33 x 10"3 03 (25)
where
E = erosion rate, ym/yr
RH = relative humidity, %
SC>2 = concentration of S02, yg/m3
63 = concentration of 63, yg/m3
Marble specimens exposed to clean air conditions produced data with con-
siderable scatter. An analysis of variance, however, showed no statistically
significant effects and mean erosion rates were not significantly different
from zero.
Although relative humidity and levels of S0£ and 03 appear to accelerate
the erosion rate of white Cherokee marble, the exposure results and calculated
erosion rates were, nevertheless, quite low. The empirical function predicts
that this particular marble will erode only 1 mm in about 300 years when
exposed to a somewhat humid (80% RH) environment containing 80 yg/m3 (0.03 ppm)
S02--equivalent to the annual mean national primary ambient air quality
standard—and 60 yg/m3 (0.03 ppm) 03. Air pollutants, however, can alter the
surface characteristics of marble and therefore reduce inherent asethetic values.
Cement
Accurate weights of conditioned cement specimens were all but impossible
to measure under the existing laboratory conditions. The reason for this was
that the specimens absorbed atmospheric moisture at an exceedingly high rate.
Since a weighing room maintained at constant relative humidity and temperature
was not available, effects research on cement was discontinued.
83
-------
SECTION XV
REFERENCES
1. Clean Air Amendments of 1970. Public Law 91 604. December 31, 1970.
2. Spence, J.W., F.D. Stump, F.H. Haynie, and J.B. Upham. Environmental
Exposure System for Studying Air Pollution Damage to Materials. U.S.
Environmental Protection Agency, Research Triangle Park, North Caro-
lina 27711. Pub. No. EPA-650/3-75-001. January 1975. 39 p.
3. Spence, J.W. and F.H. Haynie. Design of a Laboratory Experiment to
Identify the Effects of Environmental Pollutants on Materials. In:
Corrosion in Natural Environments, ASTM STP 558, American Society
for Testing and Materials. 1974. p. 279-291.
4. Salmon, R.L. Systems Analysis of the Effects of Air Pollution on
Materials. Midwest Research Institute. Contract CPA-22-69-113.
January 1970. 187 p.
5. ASTM Standard G 1-67, Recommended Practice for Preparing, Cleaning
and Evaluating Corrosion Test Specimens. In: 1970 Annual Book
of ASTM Standards, Part 31 Metals—Physical, Mechanical, Nondestruc-
tive, and Corrosion Tests, Metallography, Fatigue, Effect of Temper-
ature. 1970. p. 928-931.
6. Guttman, H. and P.J. Sereda. Measurement of Atmospheric Factors
Affecting the Corrosion of Metals. In: Metal Corrosion in the
Atmosphere, ASTM STP 435, American Society for Testing and Materials.
1968. p. 326-359.
7. Guttman, H. Effects of Atmospheric Factors in the Corrosion of
Rolled Zinc. Ibid. p. 223-239.
8. Haynie, F.H. and J.B. Upham. Effects of Atmospheric Pollutants
on Corrosion Behavior of Steels. Materials Protection and Performance.
10: 18-21, November 1971.
9. Haynie, F.H. and J.B. Upham. Correlation Between Corrosion Behavior
of Steel and Atmospheric Pollution Data. In: Corrosion in Natural
Environments, ASTM STP 558, American Society for Testing and Materials.
1974. p. 33-51.
10. Spence, J.W. and F.H. Haynie. Pitting of Galvanized Steel in
Controlled Clean Air Environments. To be published by ASTM.
11. Larrabee, C.P. and Ellis, O.B. Report of Subgroup of Subcommittee
84
-------
VII on Corrosiveness of Various Atmospheric Test Sites as Measured
by Specimens of Steel and Zinc. Proc. of American Society for
Testing and Materials. 59: 183-201. 1959.
12. Corrosiveness of Various Atmospheric Test Sites as Measured by
Specimens of Steel and Zinc. In: Metal Corrosion in the Atmosphere,
ASTM STP 435. American Society for Testing and Materials. 1968.
p. 360-391.
13. Haynie, F.H. and J.B. Upham. Effects of Atmospheric Sulfur Dioxide
on the Corrosion of Zinc. Materials Protection and Performance.
9: 35-40, August 1970.
14. Stress Corrosion Testing Methods. In: Stress Corrosion Testing,
ASTM STP 425, American Society for Testing and Materials. 1967.
p. 3-20.
15. Campbell, G.G., G.G. Schurr, D.E. Slaevikowski, and J.W. Spence.
Assessing Air Pollution Damage to Coatings. Jour, of Paint Technology.
46 (593): 59-71, June 1974.
16. Beloin, N.J. Fading of Dyed Fabrics Exposed to Air Pollutants:
A Chamber Study. Textile Chemist § Colorist. 5: 128-133, July 1973.
17. Turner, T.H. Damage to Structures by Atmospheric Pollution. Smokeless
Air fOxford, Eng.). 23: 22-26, April 1952.
18. Braun, R.C. and M.J.C. Wilson. The Removal of Atmospheric Sulphur
by Building Stones. Atmospheric Environment (London). 4(4): 371-378,
1970.
85
-------
TECHNICAL REPORT DATA
(Please read Instructions on the reverse before completing)
\. REPORT NO.
EPA-600/3-76-015
2.
3. RECIPIENT'S ACCESSION NO.
4. TITLE AND SUBTITLE
EFFECTS OF GASEOUS POLLUTANTS ON MATERIALS-
A CHAMBER STUDY
5. REPORT DATE
February 1976
6. PERFORMING ORGANIZATION CODE
7. AUTHOR(S)
8. PERFORMING ORGANIZATION REPORT NO.
F. H. Haynie, J. W. Spence and J. B. Upham
9. PERFORMING ORGANIZATION NAME AND ADDRESS
Environmental Sciences Research Laboratory
Office of Research and Development
U.S. Environmental Protection Agency
Research Triangle Park, N.C. 27711
10. PROGRAM ELEMENT NO.
1AA008
11. CONTRACT/GRANT NO.
12. SPONSORING AGENCY NAME AND ADDRESS
13. TYPE OF REPORT AND PERIOD COVERED
Inhouse
same as block 9
14. SPONSORING AGENCY CODE
EPA-ORD
15. SUPPLEMENTARY NOTES
16. ABSTRACT
This document describes a comprehensive laboratory study using specially
designed controlled environment exposure chambers to as.sess the effects of
gaseous air pollutants (sulfur dioxide, nitrogen dioxide, and ozone) on a
variety of materials. Materials included weathering steel, galvanized steel,
aluminum alloy, paints, drapery fabrics, white sidewall tire rubber, vinyl
house siding, and marble. The exposure experiment was statistically designed
using a two-level factorial arrangement to identify the environmental factors or
combination of factors, or both, that cause materials damage. Over 200
different direct and synergistic effects were examined. The study revealed that on!)
22 of the possible effects were statistically significant at better than a 95
percent confidence level. Sulfur dioxide, relative humidity, and the interaction
between them, were the main factors causing effects. A number of empirical functions
were developed that relate materials effects to various factors causing the effects.
An exceptionally good relationship was obtained for the corrosion of weathering
steel.
The lack of statistical significance that was found for the large majority of
effects that were studied is equally as important as the significant effects. As a
result a large number of material-pollutant combinations may be excluded from further
7.
KEY WORDS AND DOCUMENT ANALYSIS
DESCRIPTORS
b.IDENTIFIERS/OPEN ENDED TERMS
c. COSATI Field/Group
Experimental design Statistical
*Test chambers analysis
Air Pollution
*Sulfur dioxide
*Nitrogen dioxide
*0zone
^Degradation *Materials
14B
13B
07B
14G
11
12A
3. DISTRIBUTION STATEMENT
RELEASE TO PUBLIC
19. SECURITY CLASS (This Report)
UNCLASSIFIED
21. NO. OF PAGES
20. SECURITY.CLASS (Thispage)
UNCLASSIFIED
22. PRICE
EPA Form 2220-1 (9-73)
86
------- |