EPA-650/2-73-025
APPLICATION
OF ODOR TECHNOLOGY
TO MOBILE SOURCE
EMISSION INSTRUMENTATION
by
P.L. Levin and D. Kendall
Arthur D . Little Incorporated
15 Acron Park
Cambridge, Massachusetts 02140
Contract Number 68-02-0561
Program Element No. 1A1010
EPA Project Officer: John E. Sigsby , Jr.
Chemistry and Physics Laboratory
National Environmental Research Center
Research Triangle Park, N.C. 27711
Prepared for
COORDINATING RESEARCH COUNCIL
30 ROCKEFELLER PLAZA
NEW YORK, NEW YORK 10020
PROJECT CAPE-7-68
OFFICE OF RESEARCH AND DEVELOPMENT
U.S. ENVIRONMENTAL PROTECTION AGENCY
WASHINGTON, D.C. 20460
September 1973
-------
This report has been reviewed by the Environmental Protection Agency and
approved for publication. Approval does not signify that the contents
necessarily reflect the views and policies of the Agency, nor does
mention of trade names or commercial products constitute endorsement
or recommendation for use.
11
-------
ABSTRACT
Application of Odor Technology to Mobile
Source Emission Instrumentation
P. L. Levins and D. A. Kendall
An instrument has been developed which measures the odorant
species in diesel engine exhaust. The emissions are collected in
traps by pulling the sample through them. The traps are transported
to the instrument and the odorants eluted for analysis. Instrumental
response has been correlated with odor intensity for the species present.
The instrument represents the culmination of a series of contracts
which defined odor character and the analytical technology necessary
to measure it, both subjectively by panels of human noses and analytically.
The instrument is a liquid chromatograph which separates the collected
material into three fractions: paraffinic, "oily-kerosene" and "smoky-
burnt." The first fraction has no odor and is not seen by the ultra-
violet detector. The last two fractions contain the odorous species and
are detected. The names are indicative of the type of odor character
seen.
The report describes correlations between human response and the
instrumental response. The instrument has built-in adjustments to
account for radical changes in odorants if they occur.
Calibrations are made with the fuel used by the engine and by pure
compounds. Sampling problems have been overcome and initial data
indicates that samples may be stored for up to two weeks. If large samples
are collected, the odorants may be presented to a human panel, as well as
the instrument.
Firmcorrelations with other odor techniques have not been made, but
are under way with the cooperation of the Mobile Source Pollution Control
Program.
The report was submitted in fulfillment of contract number 68-02-0561
by Arthur D. Little, Incorporated, Cambridge, Massachusetts, under the
joint sponsorship of the Environmental Protection Agency and the
Coordinating Research Council (APRAC) CAPE-7-68. The work was completed
as of August 1, 1973.
ill
-------
ACKNOWLEDGMENT
It has been especially rewarding during this contract to pro-
ceed from the knowledge of the odorous diesel exhaust species to devel-
oping means for their measurement. We would like to thank the CAPE-7
Committee members whose guidance and contributions of engine and fuel
technology have been especially helpful:
Dr. Roderick S. Spindt - Chairman, Gulf Research and Development
Roger C. Bascom - Cummins Engine Company
Dr. Gerald J. Barnes - General Motors Research Laboratories
Fred J. Hills - Mobil Research and Development Corporation
Dr. Ernest W. Landen - Caterpillar Tractor Company
John E. Sigsby, Jr. - Environmental Protection Agency
Helmut F. Butze, National Aeronautics and Space Administration
Mr. Timothy C. Belian's efforts on behalf of the CRC office
are also gratefully acknowledged.
PROJECT TEAM
The research program at ADL has been under the overall manage-
ment of Dr. Philip L. Levins. Technical direction has been provided
jointly by Dr. Levins (analytical) and Mr. David A. Kendall (sensory).
Miss Alegria B. Caragay and Dr. James E. Oberholtzer have assumed major
responsibilities for the analytical program, and Mr. Gregory Leonardos
for the sensory evaluation.
We would also like to acknowledge the technical assistance of
Mr. James L. Stauffer, Mr. Rafael A. Cruz-Alvarez, and Mr. Ralph C.
Steeves. Odor panel members were David A. Kendall, Gregory Leonardos,
Shirley A. Raymond, Loren B. SjbstrSm and Frederick Sullivan.
IV
-------
TABLE OF CONTENTS
Page
ACKNOWLEDGMENT IV
PROJECT TEAM iv
LIST OF TABLES Vi
LIST OF FIGURES
GLOSSARY
I. SUMMARY 1
II. INTRODUCTION 3
III. SAMPLE COLLECTION 7
A. Sample Storage 7
B. Analytical Traps 9
C. Elution Solvents 13
D. Chromosorb Activation 13
E. Collection Temperatures 13
IV. DEVELOPMENT OF DATA BASE 19
A. Field Trip Samples 19
B. Fuel Composition 27
C. Correlations 27
V. ALC INSTRUMENT 37
A. Description of Method 37
B. Solvent Selections 37
C. Detector Response 41
D. ALC Instrument Description 49
VI. RECOMMENDATIONS 61
VII. REFERENCES 63
APPENDIX A - Odor Profile Measurements
APPENDIX B - Chemical Composition of the
Odor Components in Diesel Exhaust
APPENDIX C - Sample Collection Procedures
-------
LIST OF TABLES
Table No. Page
1 Effect of Sample Storage and Preparative Trap Reuse
on Diesel Exhaust Odor and Analytical Composition 8
2 Analytical Trap (AC) Exhaust Analysis Results 10
3 Diesel Exhaust Samples From No. 1 Fuel:
Comparison of Analytical and Preparative Samples 11
4 Diesel Exhaust Samples from No. 2 Fuel:
Analytical and Preparative Samples 12
5 Comparison of Pentane or Cyclohexane AC Trap Elution .. 14
6 Blank Background Associated with Analytical Traps 15
7 Company A Engine Condit ions 20
8 On-Site Odor Observations at Company A 21
9 Company B Test Engines and Odor Observations 23
10 Company C Test Engines and Odor Observations 24
11 Summary of Analytical and Odor Data from Field Trips .. 25
12 Correlation Check with Company B Exhaust Samples 26
13 Fuel Properties 28
14 Fuel Variables Analytical and Odor Data 29
15 Odor and Analytical Correlation of Preliminary
Test Data 31
16 Statistical Analysis Summary 32
17 Summary Odor-LCO Correlation Data Set 34
18 ALC LCA Fuel Response Characteristics 42
19 Dependence of ALC Detector Response for LCO 45
20 Reference Compound ALC Detector Response 48
VI
-------
LIST OF FIGURES
Figure No.
1 Relationship of Diesel Exhaust Odor Components 4
2 Temperature Difference Between Preparative and
Analytical Traps During Samples Collection 17
3 Comparison of Exhaust and TOE Odor Intensities 30
4 Current Summary Correlation of Exhaust Odor
Intensity and LCO 35
5 Typical ALC Chromatogram of a Diesel Exhaust
Sample 38
6 ALC Chromatogram of Diesel Exhaust with the
Cyclohexane-Isopropanol Solvent System 40
7 Fuel ALC Response 43
8 Relation of ALC Response to Aromatic Content 44
9 ALC LCO Detector Response 46
10 ALC Instrument Overall Configuration 50
11 Diesel Odor Analysis Instrument 51
12 Rear View of Diesel Odor Analysis Instrument 52
13 Relationship of Timing Control Logic and
Operating Components 53
14 ALC Logic Timing Diagram 55
15 UV Detector Analog Computer Signal Processing 57
16 ALC Instrument Analysis of Diesel Exhaust 59
Vli
-------
GLOSSARY
The following terms have been used frequently in the text and
are summarized here with their definitions, for the convenience of the
reader.
TOE - Total Organic Extract, the total organic exhaust
species isolated from the sample collection by
solvent extraction.
LCP - The Paraffin (odorless) exhaust fraction isolated
from the preparative Liquid Chromatography procedure.
LCA - The Aromatic (oily-kerosene) exhaust fraction iso-
lated from the preparative Liquid Chromatography
procedure.
LCD - The Oxygenate (smoky-burnt) exhaust fraction
isolated from the preparative Liquid Chromato-
graphy procedure.
TIA - Total Intensity of Aroma. Frequently tabulated
as the value computed from the dose/response
data at a 1000/1 dilution (1 £/m3) of exhaust
or odor fraction.
yg/£ - Concentration of exhaust species reported as yg
per i of exhaust.
mg/k& - Equal to yg/Jl and mg/m3.
£/m3 - Liters of exhaust, or equivalent amount of an
exhaust odor fraction, per m3 of air. Reciprocal
of dilution.
ALC - Analytical Liquid Chromatography - the instru-
mental odorant measurement method.
UV - Ultraviolet absorption. The detector system
used in the ALC recording optical density at
254 nm.
OD - Optical density. The UV detector output signal.
Area/yg - Integrated ALC UV detector response per yg of
sample; units of OD sec/yg.
CHROM - Sampling coding used for preparative size
(500 H) Chromosorb 102 sample collection
traps.
AC - Analytical size sample traps (10 - 50 £).
ACS - As above, different geometry. Method now
used with instrument system.
viii
-------
I. SUMMARY
In the previous studies on this project C1 **), diesel exhaust
odor has been characterized as consisting of oily-kerosene and smoky-
burnt odor groups. Of the two, the smoky-burnt is the more intense
and most characteristic of the combustion process. These oily-kerosene
and smoky-burnt odor groups are found respectively in the aromatic (LCA)
and oxygenate (LCD) portions of separated fractions of collected diesel
exhaust odor. Odor and chemical correlation studies have shown that
the odor intensity of diesel exhaust is a logarithmic function of the
concentration of the LCO fraction in the exhaust. The results of these
correlation studies have been used for the development of an instru-
mental method for measurement of the odor intensity of diesel exhaust.
The instrumentation is now available for application to studies of means
to decrease exhaust odor levels. The instrumental method is based on
analytical liquid chromatography using an ultraviolet detector for mea-
surement of the odorous fractions.
Several detailed studies were necessary before development of
the final version of the analytical liquid chromatograph (ALC) instru-
ment. Cooperative participation of the research laboratories of three
diesel engine manufacturers provided chemical composition and odor data
from a much wider range of fuel and engine conditions than has been pre-
viously available. From the data derived from these and the preceding
studies, the final correlation between diesel exhaust odor intensity
(TIA*) and LCO smoky-burnt exhaust fraction abundance was developed as
TIA = 1.0 + 1.0 log LCO
where the concentration of LCO is expressed as Vg/£ of exhaust, and the
TIA value is for the odor intensity at a 1000/1 dilution (1 £/m3) of
exhaust.
New analytical-size sample collection traps were designed to
be compatible with the ALC instrument sample size requirements and al-
low the collection of sufficient sample for analysis in 1 - 5 minutes
depending on the odor level. Sample storage experiments showed col-
lected samples of exhaust to be stable for about 12 days. Shipment
from remote points to a central laboratory for analysis is thus a re-
ality and was practiced regularly during the cooperative study.
* Total Intensity of Aroma - See Appendix A for details.
-1-
-------
Detector response of the ALC was studied as a function of fuel
type, using four fuels covering a range from kerosene to a high aromatic
No. 2 fuel. The LCA response is fuel composition (% aromatics) depen-
dent, while the LCO response is independent of composition. Calibration
standard solutions have been selected for use with the ALC system.
The ALC instrument has been designed to be as reliable and con-
venient to use as possible. It may be operated manually for detailed
research studies or operate completely automatically for routine analyses.
The system was constructed from proven components, and the electronic
components for detector output signal processing are all solid state
circuits. Support requirements for use of the instrument, such as sol-
vents, have been simplified to a convenient two solvent system. The
ALC provides front panel selectable recorder output in terms of optical
density (OD), yg of LCA and LCO, or odor intensity (TIA) of exhaust.
—2—
-------
II. INTRODUCTION
The chemical species responsible for the characteristic diesel
exhaust odor have been identified in the previous effort on this pro-
gram^1"1*). The effort during the most recent of these studies ^ was
to obtain quantitative means of collecting and analyzing the odorous
species in order that an instrumental method could eventually be devel-
oped. Our research program continually emphasizes the interrelationship
between the sensory measurement by experienced odor chemists and the
analytical chemical measurements while developing the measurement schemes.
Some details of the previous work which are especially relevant
to the work reported here are included in Appendices A-C. The basic
odor measurment techniques and new dose/response odor measurement method
are described in Appendix A. The chemical composition and odor assign-
ments of the two major exhaust odor groups, smoky-burnt and oily-kerosene,
are summarized in Appendix B. Appendix C describes the details of the
diesel exhaust odor collection methods used in these studies.
The details of our studies of typical diesel exhaust odors may
be summarized by describing diesel exhaust as consisting of two princi-
pal groups, oily-kerosene and smoky-burnt, each of which contributes
significantly to diesel exhaust odor. The smoky-burnt odor is normally
the most intense of the two odor groups and most characteristic of odors
associated with the combustion process. A new procedure, which calculates
the exhaust odor intensity at a 1000/1 dilution from the dose/response
data, is used in place of the earlier procedures of reporting TIA values
at a single dilution.
Methods for the quantitative collection of the diesel exhaust
odors using Chromosorb 102 are detailed in Appendix C. The total organ-
ic extract (TOE), isolated from the sample traps by solvent elution, may
be resolved into three functionally distinct chemical groups using liquid
chromatography (LC) methods. Species with volatility equal to or greater
than pentane would not be detected by these procedures, but they have
much less odor than the higher molecular weight species.
This approach separates the TOE sample into 1) an odorless par-
affin fraction (LCP), 2) an aromatic fraction (LCA1 containing the oily-
kerosene odor group, and 3) an oxygenate fraction (LCO) containing the
smoky-burnt odor group. The overall relationship of these samples and
fractions is shown in Figure 1.
The previous work CO had shown that the smoky-burnt LCO exhaust
fraction was the most significant odor group and indicated that it was
-3-
-------
Diesel Exhaust
Collection on
Chromosorb 102
and Elution
TOE
(Total Organic Extract)
Liquid Chromatography
Fractionation
LCP
Paraffins
No Odor
LCA
Aromatics
Oily-Kerosene
LCO
Oxygenates
Smoky-Burnt
FIGURE 1 RELATIONSHIP OF DIESEL EXHAUST ODOR COMPONENTS
-A-
-------
possible to measure the total diesel exhaust odor intensity (TIA, see
Appendix A) by measurement of the abundance of the total LCO group.
The LCO measurement was made using an analytical liquid chromatography
(ALC.) method utilizing an ultraviolet absorption detector. The prelim-
inary studies indicated that the exhaust odor intensity could be measured
by a relationship of the form
TIA = a + b log LCO
where the odor intensity is the value computed at a diluted exhaust con-
centration of 1 H exhaust/m3 of dilution air (1 £/m3), and the LCO con-
centration is reported in ug/& (or mg/m ) of exhaust.
The purpose of the research described in this report was, first,
to verify our preliminary observations of the TIA-LCO correlation by
studying exhaust generated over a wide range of conditions, and, second,
to develop a final (automated) instrument which could be used in any
diesel odor research laboratory.
-------
-------
III. SAMPLE COLLECTION
The basic sample collection procedures using Chromosorb 102
have been reported ^', and the procedure is summarized in Appendix C.
A preparative size trap containing 10 g of Chromosorb is used to collect
sufficient sample (500 &) for odor and analytical studies; these are
coded CHROM. The analytical size trap contains about 1 g of Chromosorb
and is used to collect 10 - 50 £ of exhaust for instrumental analysis;
these traps are coded AC and ACS. Several details remained to be examined,
such as sample storage characteristics and collection temperature data.
Since development of the new ALC instrumental method required small
amounts of sample (0.1 - 1 £), smaller and simpler analytical scale
traps could be used which were more economical and faster to use. The
results of these studies are described in this section.
A. Sample Storage
The preparative-scale exhaust sample collection traps containing
10 g of Chromosorb 102 in a 1" x 5" cylinder were used to determine the
length of time a sample could be stored without loss of information.
We also examined the possibility of regenerating the absorbent in situ
by percolating pentane/methylene chloride/pentane through the trap and
oven drying at 70°C. The results of these studies are summarized in
Table 1.
*
Samples were obtained from the 4-71 engine operating at 33%
load, 1800 rpm, with N-60 injectors and No. 1 diesel fuel. The aromatics
(LCA) and oxygenates (LCD) concentration were obtained from the prototype
analytical method. Odor data represent the least squares summary of the
semi-logarithmic dose/response of each sample. One can see that the data
show no change after 7 days of storage of the sealed sample on the labo-
ratory bench. After 14 days there was a measurable loss in the quantity
of the LCD fraction and in the odor intensity. We estimate that samples
can be stored for 7 to 10 days without difficulty and thus can be readily
shipped between laboratories.
The sample collection tube from Chrom 56 was reactivated in
situ followed by collection of Sample 56A, further in situ reactivation
and collection of Sample 56B. The traps were reactivated by flushing
with methanol and pentane and oven drying at 70°C. The LCO values for
56, 56A and 56B are nearly the same, but the reactivated traps show a
slight decrease in collection efficiency. Since all Chromosorb must be
conditioned before use, we believe it would be best to repack the traps
each time. Repacking the traps also provides the opportunity to avoid
undetected leaks and particulate contamination. Chromosorb removed from
the used traps may be adequately regenerated using the same procedure as
for new material.
Detroit Diesel Allison Division of General Motors Corporation
4-71 Diesel Engine, Model No. 4154N
-7-
-------
Table 1
Effect of Sample Storage and Preparative
Trap Reuse on Diesel Exhaust Odor and Analytical Composition
ALC/UV Data Odor Data
Sample Code3 New or Reused Days Storage LCA LCO TIA at 1 &/m3 Slope
Chrom 55
Chrom 57
Chrom 58
Chrom 56
Chrom 56A
Chrom 56B
Trap Condition
N
N
N
14
0
R
38 4.4
49 8.0
48 6.5
40 6.2
40 5.6
40 5.5
1.38
1.89
1.87
1.90
2.09
1.70
1.74
1.83
1.66
1.57
2.24
1.66
a. 500 liters of exhaust were collected in each case
-8-
-------
B. Analytical Traps
The new ALC instrumental method described in Section V requires
the analysis of the equivalent of only a 0.1 - 1 2. aliquot of the col-
lected sample of exhaust depending on the odor level. It was, there-
fore, appropriate to examine other sample collection trap configurations
which could be used to collect only enough sample for the ALC analysis
and thus be less expensive, easier to use, and permit shorter sample
collection times.
Two different sets of trap dimensions were examined, each using
about 1 g of Chromosorb 102. Both are described in detail in Appendix
C. The first was constructed from 5-inch lengths of V OD stainless
steel tubing with a I/A" - 1/8" Swagelok reducing union at each end.
These traps (coded AC) provided sample collection rates of 1 - 2 £/min.
The second type, having the same volume but a larger cross-sectional
area, was fabricated from 2-inch lengths of 3/8" OD tubing with a 3/8" -
1/8" Swagelok reducing union at each end. Samples were normally col-
lected at 10 £/min with these traps (coded ACS). These latest ACS traps
are the preferred configuration, since they allow minimum collection
times of 1 - 5 minutes per sample to provide sufficient material to ana-
lyze exhausts covering a wide range in emissions.
A number of experiments were run with the AC analytical traps
to compare the effects of sample storage and in situ reactivation com-
parable to the studies with the preparative scale traps. These results
are summarized in Table 2. There is no effect observed for storage up
to 12 days. Reactivation is again seen to lower the collection effi-
ciency, but to a reproducible level. For reasons stated before, reacti-
vation in situ is not recommended.
The collection efficiency of the various traps was compared by
collecting several sets of samples from an engine run on both No. 1 and
No. 2 fuel. The results are given in Tables 3 and 4 using 10-liter ex-
haust collections for the analytical traps and the standard 500-liter
size sample for the preparative Chrom traps.
There does not appear to be any significant difference between
the various analytical trap configurations. There is, however, a con-
sistent difference between the analytical and preparative traps with
lower LCD values observed from the preparative traps. We believe this
may be due to the prolonged heating of the sample during the one-hour
collection period for the preparative samples and some reaction losses.
Such effects indicate that greater attention must be given to sampling
temperature effects. Some data on this matter are given later in this
section, but a more detailed definition of sample collection temperature
requirements are still needed.
-9-
-------
Table 2
Analytical Trap (AC) Exhaust Analysis Results
AC Trap Condition
Fresh traps, not stored
Stored for 12 days
Exhaust Fraction
LCA (yg/A) LCO
30 (a = + 5) 6.1 (o
28 5.8
+ 1.1)
Reactivated, not stored 22 (a = + 4) 5.4 (a = + 0.6)
a. Analysis of 0.5 H from 15 H of collected sample.
-10-
-------
Table 3
Diesel Exhaust Samples From No. 1 Fuel:
Comparison of Analytical and Preparative Samples
Sample Code
AC-54
AC-55
AC-56
ACS-12
ACS-13
ACS-14
Chrom 88
Chrom 89
Chrom 90
(1)
Average AC
Average ACS
LCA
46
47
49
56
52
56
46
45
48
-ALC/UV Data in
48
54
Average Preparative Samples 46
LCO
6.0
5.6
5.6
6.1
5.3
6.0
3.9
5.1
3.9
5.7
5.8
4.3
(1) All analytical samples were 10 liters, prep samples
were 500 liters of exhaust.
-11-
-------
Table 4
Diesel Exhaust Samples from No. 2 Fuel:
Analytical and Preparative Samples
Sample Code
(1)
LCA
ALC/UV Data in yg/£
LCD
AC-57
AC-58
AC-59
ACS-15
ACS-16
ACS-17
CHROM 91
CHROM 92
CHROM 93
Average AC
Average ACS
32
30
28
49
43
42
30
45
54
68
29
Average preparative samples 50
7.4
7.0
6.8
8.6
8.1
7.4
5.3
6.1
4.5
7.1
8.0
5.3
(1) All analytical samples were 10 liters, prep samples
were 500 liters of exhaust.
-12-
-------
C. Elution Solvents
The pentane used to elute the preparative scale traps is still
preferred for those traps, since the sample is primarily used for odor
and preparative liquid chromatography studies. Pentane provides minimum
interference for those analyses. The optimum solvent choice for the ALC
instrument is now cyclohexane/isopropanol, and it would be desirable to
keep the number of different reagents required for the ALC analysis to
a minimum. A series of 10 £ AC samples were collected from a No. 2 fuel
exhaust and eluted with either pentane or cyclohexane. The data shown in
Table 5 indicate no difference between the solvents and cyclohexane thus
becomes the solvent of choice for the analytical traps.
D. Chromosorb Activation
A small amount of LCD is sometimes observed in analyzing blank
analytical traps. Some earlier data had indicated that this could be
due to decomposition during oven drying after solvent washing. Several
experiments were run comparing results from oven and air dried Chromosorb.
The last set of results in Table 5 show that the same collection efficien-
cies are obtained with either oven or air dried (spread on a petri dish
and allowed to stand).
Analysis of blank AC traps was also done using pentane or cyclo-
hexane and oven or air drying as reported in Table 6. Clearly, the air
drying procedure leads to lower background levels and is the preferred
procedure.
E. Collection Temperatures
All samples studied to date have been collected through a sam-
ple train containing a heated particulate filter and heated sample lines.
A detailed study of the temperature conditions required for optimum sam-
ple collection has not, however, been carried out. The first approxi-
mation would be that the same condition'-' as used for diesel exhaust hydro-
carbon analysis would be appropriate. The Chromosorb sample tube must,
however, be maintained at a low temperature or substrate decomposition
would be a problem.
We have recently collected both preparative and analytical
traps so that the normally encountered temperatures could be measured.
The actual temperature in the sample trap was measured by imbedding
thermocouples in the Chromosorb. For the preparative trap, thermocouples
were placed near the entrance and exit of the trap. A single thermo-
couple was placed near the entrance of the analytical (AC) trap. The
4-71 engine was run at 33% load, 1800 rpm, No. 2 fuel, N-60 injectors,
and the average exhaust temperature was about 204 C. The heated parti-
culate filter was observed to be 215°C for each sample.
-13-
-------
Table 5
Comparison of Pentane or Cyclohexane AC- Trap Elution
„(!)
Solvent and
Sample Code Chromosorb Condition LCA LCO
(2)
AC-60 Pentane, oven dried 31 5.0
AC-62 30 5.0
AC-64 30 5.5
AC-66 29 5.2
AC-68 33 6.6
average ........ 31 5.4
(2)
AC-61 Cyclohexane, oven dried 33 5.0
AC-63 32 7.3
AC-65 30 6.0
AC-67 32 5.5
AC-69 32_ 5.5.
average ........ 32 5.9
AC-70 Cyclohexane, air dried ^ ' 32 6.3
AC-71 30 5.5
AC-72 32 6.4_
average ........ 31 6.1
(1) All values have been corrected for the blank
absorption.
(2). The prewashed and dried Chromosorb was activated
at 85°C for 16 hours.
(3) The same batch of prewashed Chromosorb has been
air-dried only.
-14-
-------
Table 6
Blank Background Associated with Analytical Traps
(1)
ug/10 vt
(3)
Solvent
Pentane
Pentane
Pentane
Cyclohexane
Cyclohexane
Chromosorb Activation LCA
24 hours at 85° C 0.5
air dried 0
16 hours at 85°C 0.3
16 hours at 85°C 0.2
air dried 0
LCD
0.05
0
0.08
0.15
0.10
(1) AC traps packed with oven or air dried
Chromosorb and extracted with 1.0 mX, of
pentane or Cyclohexane.
(2) After solvent washing (See Appendix C)
(3) 10 pit is the maximum solvent volume normally
used in a sample analysis.
-15-
-------
The results are shown in Figure 2. Sample flow rates were about 10 Jl/min
for the preparative trap and 2 fc/min for the AC trap.
The preparative trap reached 100°C after 10 minutes at which
time the temperature stabilizes, possibly due to the controlling effect
of water vaporization. The analytical trap reaches a temperature of
only 60°C in the time required to collect a normal sample. The higher
temperature observed in the preparative trap could explain the occasional
observation of lower LCA and LCD values from the prep traps. Lower values
could be due to volatility losses (although the exit FID-THC level is low)
or reactivity losses on the surface.
-16-
-------
a
ID
2
O
CM
c
j
i
F
CM
Ul
oc
a
0°
-17-
-------
-------
IV. DEVELOPMENT OF DATA BASE
A. Field Trip Samples
A series of field trips was initiated to explore several fac-
tors important to the odor analysis and instrument development studies.
It was necessary to be sure that samples could be easily collected at
other facilities using the same basic procedures as in our laboratories.
These trips also provided a convenient means for checking any unusual
transportation effects on the samples such as might be incurred in air
travel. Most importantly, the trips provided the most efficient means
of analyzing exhaust from a wide range of engine and fuel conditions,
especially low emission diesels. Previous results had been restricted
to the range of conditions possible with the 4-71 engine.
The research laboratories of three diesel engine manufacturers,
referred to as Companies A, B and C, participated in the study. In each
case a visit was made to the Company by an ADL odor and analytical chem-
ist to conduct preliminary odor surveys and collect samples. Companies
A and B later collected samples using their own personnel and shipped
them to ADL for analysis. Both preparative scale, 500.£, (CHROM) samples
and analytical scale, 10 SLt (AC) samples were collected.
At Company A an odor survey was made at the roof vents of many
test cells, and several were then selected for study. At Companies B
and C two different sets of engines were identified for study. The ob-
servations made at each of these laboratories and the results obtained
are detailed in the following discussion.
1. Company A
The engines selected for study were based initially on a roof-
top vent odor survey and later selected for studies of specific effects.
These engines were all supplied with No. 2 diesel fuel and were most re-
presentative of the different exhaust odor types selected from about 30
exhaust conditions. The engines or conditions studies are given in Table
7, and the odor observations are given in Table 8.
At Company A the exhaust from several engines had pungent alde-
hydic odor character notes. Several of the stack emissions appeared to
have dominant character notes which we do not observe as such from the
4-71 engine but have seen in fractions such as naphthanate, MCP and musty.
One engine with high smoke was not too high in odor — another was pro-
bably most similar to the exhaust from the 4-71N engine.
In the afternoon examination, some of these characteristics ap-
peared to have changed and, therefore, may only represent transient or
sampling variations. A "cold start" gave the highest odor intensity;
-19-
-------
Table 7
Company A Engine Conditions
Engine
Al
A2
A3
A4
rpm
2100
3300
2100
2100
Sample
THC(ppm C) CHROM
65a 47
170a 48
65a 49
a. Estimated from previous results
b. Endurance test at end of run (200 hours)
before oil change
c. Same as above just after oil change
51
A5 1500 35 72, 73
A6b 1000 - 74, 75
A6C 1000 - 76, 77
-20-
-------
Table 8
On-Site Odor Observations at Company A
Engine/Sample
Al/CHROM 47
Roof Vent
TIA 1
MCA (burnt sweet)
Burnt oil
Sour
Pungent
Odor
At Engine
TIA 1
Burnt oil
Sweet
Particulate
Hot metal
A2/CHRQM 48
A4/CHROM 51
TIA 2%
Tarry Asphalt
Burnt oil
(not smoke)
TIA 1
Oily
Kerosene
Burnt
TIA 2-2^
Sour
Tarry
Smoky
Pungent
TIA 1
Oily
Anisole
Burnt sweet
A3/CHRQM 49
TIA 2
Sour (green oxidized oil)
Pungent
Burnt sweet
TIA 2
Smoke
Tarry Asphalt
Sour
Anisole
-21-
-------
all of the odor character notes one could think of were observed in a
constantly changing sequence.
The quantitative odor and analytical data obtained for these
samples are presented later in a summary of all the field trip data
(Table 11).
2. Company B
The engines studied at Company B represent a current production
model and a new low emissions engine. The engine conditions and odor ob-
servations are summarized in Table 9. Samples 80-83 were shipped later
in the program and provided an excellent opportunity to compare predicted
TIA odor values from the ALC instrument with the observed odor panel val-
ues, (see Table 12).
3. Company C
Two engines of different displacement were studied, and dupli-
cate preparative and analytical samples were collected at Company C.
The engines were running on a No. 2 diesel fuel. A description of the
engines and summary of the analytical and odor data are given in Table
10.
The odor characteristics of the engine Cl exhaust were similar
to those observed with our 4-71; primarily burnt smoky, oxidized oil,
naphthenate and a low level kerosene. A low intensity breakthrough of
burnt oily was observed at the end of the sample collection. The C2
exhaust odor was primarily burnt sooty with some musty naphthenate. The
quantitative dose/response odor studies on collected samples show the two
exhausts to be quite similar in odor intensity.
4. Analytical and Odor Summary
The quantitative analytical data obtained from the three com-
pany's samples and the quantitative dose/response odor data are sum-
marized in Table 11. The primary purpose of acquiring these data is
to develop the correlations between measurable exhaust species (LCO)
and odor intensity. The set of experiments did also show, however, that
samples can be reliably collected under a variety of conditions for analy-
sis and that other personnel quickly learn the sampling methodology.
These data are treated in detail in the correlation section.
The last set of samples from Company B provided at least one
opportunity to test the correlation with new samples which were not in-
cluded in the data set from which the correlation was developed. Anal-
ytical and odor data for samples collected from a B3 engine at two load
conditions are summarized in Table 12. The duplicates agree quite well,
and the comparison between the observed 1 £/m3 TIA odor levels and those
calculated from the LCO values (TIA = 1.0 + 1.0 log LCO) are in excellent
agreement.
-22-
-------
Table 9
Company B Test Engines and Odor Observations
Engines
Engine rpm ppm C Sample
Bl 2200 - 53, 54
B2 2800 700 50, 52
B3 2200 (124 BHP) - 80, 81
B4 2200 (2.2 BHP) - 82, 83
Odor Observations
Engine Odor at Engine
Bl TIA 1
N0x - pungent
Burnt oil 1
Tarry %
B2 TIA 2-2*2
Smoky burnt 2
Oil (ox) 1
Kerosene 1%
Pungent
(like 4-71)
-23-
-------
Engines
Table 10
Company C Test Engines and Odor Observations
Engine
Cl
Cl
C2
C2
Load
50%
50%
50%
50%
rpm
2100
2100
2100
2100
THC(ppm C)
30
30
95
95
Sample
Chrom 68, 69
AC 30, 31
Chrom 70, 71
AC 32, 33
Odor Observations
Engine
Cl
Odor at Exhaust
Burnt smoky 2
Oxidized oil l
Naphthenate ^
Kerosene ^
C2
Burnt sooty 1
Musty naphthenate 1
Oxidized oil h
-24-
-------
Table 11
Summary of Analytical and Odor Data from Field Trips
yg/l. Odor
Company
A
A
A
A
A
A
A
A
A
B
B
B
B
C
C
C
C
C
C
C
C
Sample
47
48
49
72
73
74
75
76
77
50
52
53
54
68
69
AC 30
AC 31
70
71
AC 32
AC 33
Engine
Al
A2
A3
A5
A5
A6a
A6a
A6b
A6b
B2
B2
Bl
Bl
Cl
Cl
Cl
Cl
C2
C2
C2
C2
LCA
1.8
8.5
7.2
1.5
1.5
3.5
2.5
2.0
1.5
72
64
2.8
3.6
3
4
5
4
16
18
7
8
LCD
0.8
5.9
3.8
1.5
2.0
2.0
2.0
2.0
1.5
9.8
9.5
1.4
1.7
1.5
3.0
3.0
2.5
4.0
3.5
3.0
3.0
•TIA (1 £/m3)
0.9
1.7
1.4
1.3
1.4
1.5
1.3
1.4
1.1
2.1
2.1
0.6
1.1
1.5
1.7
-
-
1.8
1.6
-
-
Slope
1.25
1.5
1.7
1.6
1.9
1.6
1.6
1.5
1.5
1.7
1.3
1.1
1.7
1.8
1.7
-
-
1.6
1.7
-
-
a aged oil
b fresh oil
-25-
-------
Table 12
Correlation Check with Company B Exhaust Samples
Sample
CHROM
80°
81C
82d
83d
Exhaust THC yg/£
(ppm C) LCA LCO
16
15
53 12
69 11
3.8
3.9
7.0
7.8
Predicted
1.6
1.6
1.8
1.9
Observed
1.6
1.3
1.8
1.8
a. TIA at 1 2,/m3
b. from TIA = 1.0 + 1.0 log LCO
c. B3 engine, 2200 rpm, 124 B.H.P.
d. B3 engine, 220 rpm, 2.2 B.H.P.
-26-
-------
B. Fuel Composition
From several isolated earlier experiments, it appeared that
fuel composition was not a major variable in exhaust odor intensity, but
the ALC data did not clearly indicate whether there were exhaust LCD frac-
tion response variations due to differences in the initial fuel aromatic
hydrocarbons.
A series of experiments were carried out with fuels of varying
aromatic content and boiling point to provide exhaust odor fractions and
test ALC response sensitivity to this variable. For this series, a new
low aromatic kerosene was obtained. The properties of this and the other
fuels are given in Table 13. The four fuels making up the test sequence
were the kerosene (Kero), No. 1 diesel, No. 2 Midwest, and a 50/50 (V/V)
blend of the No. 1 and No. 2.
The results from these runs are given in Table 14. As found
previously, the exhaust odor levels for each are essentially the same.
These samples provided a convenient opportunity to test the adequacy of
the collected TOE sample to represent the odor intensity of the original
exhaust. The two sets of data are clearly identical. This can be seen
more clearly in Figure 3 where the TIA level of the exhaust and TOE have
been compared at two concentration levels.
The ALC response characteristics will be treated in Section V
and the correlatable data in the next section. The exhaust hydrocarbon
levels are seen to vary as expected (Ford, et al, ref. 5) with the 50/50
mix perhaps a little high.
C. Correlations
The correlation of exhaust odor intensity data with measurable
emissions, and the interrelationships of these measured emissions, has
been carried out in two phases. Many interrelationships were initially
examined using some of the new field data and the previous 4-71 emissions
data. The quantity of LCD (expressed logarithmically) was seen to be the
best measure of exhaust odor intensity. Then, when all of the field data
and data on the fuel variables were available, these were treated as a
group to derive the final TIA-LCO relationship, believed to best represent
exhaust odor intensity with the data available at this time.
The preliminary data set for which data on odor intensity, LCA,
LCD and THC were all available is listed in Table 15. The last five data
points come from the 4-71 engine — the others from the field trip samples.
The data were analyzed for the correlations indicated in Table 16 using
the linear regression method.
The log LCO value is seen to be the best predictor of exhaust
TIA by virtue of a better standard deviation (a) and goodness of fit (r2)
but also and probably more importantly, by having a slope sensitivity (b - 1.24)
-27-
-------
Table 13
Fuel Properties
roperty
Gravity, °API
% Aromatics by FIA
Distillation (°F)
Initial Pt
10%
50%
90%
End Pt
Kerosene
41.6
15.5
326
374
418
472
524
No. 1 Diesel
38.7
20.8
364
396
436
497
524
Midwest
No. 2
34.1
34.7
372
418
506
587
625
-28-
-------
Table 14
Fuel Variables Analytical and Odor Data
Odor Response
Exhaust TOE
CHROM (2) (1) (i)
Sample Fuel THCV ' LCA ICO TIAV ; Slope TIA^ ; Slope
60 No. 1 400 39 4.5 1.8 1.8 1.7 1.4
61 No. 1 400 36 4
62 Kero 500 42 4 1.6 1.5 1.6 1.6
63 Kero 530 35 4
64 No. I/No. 2 420 60 6 1.7 1.7 1.9 1.7
65 No. I/No. 2 420 54 5
66 No. 2 290 55 6 1.9 1.8 1.7 1.4
67 No. 2 300 53 6
(1) TIA at 1 2,/m3
(2) Exhaust hydrocarbon level, ppm C
-29-
-------
q
CM'
V)
LU
CO
UJ
cc
O
Q
O
Q
<
X
LU
LL
O
z
O
CO
LU
CC
D
q
CN
(0
-C
X
LU
-30-
-------
Table 15
Odor and Analytical Correlation of Preliminary Test Data
THC
TIA (1 2./m3)
0.87
1.73
1.45
2.12
2.13
0.60
1.12
1.67
2.09
1.70
1.38
1.89
1.87
2.15
1.45
1.5
1.6
1.5
LCD
0.8
5.9
3.8
9.8
9.5
1.4
1.7
6.2
5.6
5.5
4.4
8.0
6.5
14
7
5
6
6
LCA
1.8
8.5
7.2
72
64
2.7
3.6
40
40
40
38
39
48
210
39
16
38
_
(ppm (
65
170
65
700
700
40
40
660
775
800
700
750
800
3,000
620
300
550
590
-31-
-------
Table 16
Statistical Analysis Summary
y = a + bx
y
Exhaust TIA
Exhaust TIA
Exhaust TIA
THC
THC
X
Log
Log
Log
LCO
LCA
THC
LCO
LCA
a
0.75
0.69
- 0.09
- 389
5.0
b
1.24
0.68
0.66
171
13.8
r
0.
0.
0.
0.
0.
2
79
74
64
66
96
0.
0.
0.
384
138
a
195
225
256
-32-
-------
much closer to the slope value of 1.6 normally observed in the dose/re-
sponse studies of TOE or exhaust samples. THC is the poorest predictor
of TIA.
As a corollary to the odor correlations, the THC and LCA values
are seen to have a good correlation (r2 = 0.96), while there is a poor
relationship between THC and LCD (r2 = 0.66).
Finally, all of the data available to date have been combined
to examine the precision of the exhaust TIA-LCO correlation. The data
listed in Table 17 represent the average of results obtained for each
of the unique engine or variable test conditions. In most cases a data
pair represents the average of two to four separate samples. In some
cases (Al, A2, A3) only one sample was available.
The new correlation shown in Figure 4 is a marked improvement
over the former data. The correlation is:
TIA = 0.99 + 1.00 log LCO
r2 = 0.996
2cr = 0.32
The + 0.32 TIA 95% confidence limits are now better than normally observed
(0.4) in the odor observation itself.
-33-
-------
Table 17
Summary Odor-LCO Correlation Data Set
Engine or Variable TIA LCO _(ug/&)b
14
6.3
5.0
6.0
6.0
0.8
5.9
3.8
1.6
9.6
4.2
4.0
5.5
6.0
2.2
3.8
1.8
2.0
1.8
a. All odor observations from TOE samples in dynamic
chamber — average values computed from the least
squares dose/response relationship at a concentration
of 1 £/m3.
b. Averages of the separate observations.
c. Single values.
-34-
4-71 S-60
4-71 N-60
4-71 LSN-60
4-71 No. 2
4-71 Soltrol
Al
A2
A3
Bl
B2
4-71 No. 1
4-71 Kero
4-71 50/50
4-71 No. 2
Cl
C2
A5
A6
A6
2.15
1.84
1.5
1.6
1.5
.87°
1.73°
1.45°
0.86
2.12
1.8
1.6
1.7
1.9
1.6
1.7
1.35
1.40
1.25
-------
3 r~
TIA
0.5
1.0
5.0 10
LCO
50
100
FIGURE 4 CURRENT SUMMARY CORRELATION OF EXHAUST ODOR INTENSITY AND LCO
-35-
-------
-------
V. ALC INSTRUMENT
A research version of an analytical liquid chromatograph (ALC)
to measure diesel exhaust odor was assembled during the previous contract.
During the current period, the research model was converted to a prototype
version for more extensive evaluation of the data generated and detailed
definition of hardware and electronics requirements. An instrument model
has been constructed which will enable engine and fuels research labora-
tories to pursue diesel odor studies in an efficient manner. It should
be remembered that, by definition, instruments cannot measure odor —
only the olfactory sensors can do that. The instrument measures the
abundance of odorous species (odorants) which are in turn related to the
odor through the correlations generated in this study.
A brief review of the basic method is presented first to assist
the reader. In addition to the basic odor and correlation studies al-
ready discussed, several other studies required for final design of the
ALC instrument were completed. These will be presented next followed by
a description of the ALC instrument.
A. Description of Method
The basic measurment method described in detail later relies
on liquid chromatographic methods to separate the aromatic (LCA) oily-
kerosene and oxygenate (LCD) smoky-burnt odor groups for subsequent
measurement by an ultraviolet (UV) detector. A portion of the TOE sam-
ple eluted from a sample trap is injected onto a silica type (Corasil II)
chromatographic column and the paraffins (LCP) and aromatics (LCA) are
eluted rapidly in the hydrocarbon solvent. The polar oxygenate (LCO)
fraction is retained until a slug injection of alcohol into the solvent
stream elutes the LCO fraction as a separate peak. The odorant peaks
are detected by a UV detector sensitive to 254 run radiation. Since the
paraffin fraction does not absorb any UV at this wavelength, only the
LCA and LCO fractions are detected. A typical chromatogram is shown in
Figure 5.
The magnitude of the detector response is proportional to the
quantity of LCA and LCO so that this value may be used to compute exhaust
concentration and then odor intensity.
B. Solvent Selections
Most of the preliminary ALC development work has been done using
a methylene chloride-hexane/methanol solvent combination. Studies were
initiated to find a simpler solvent system for several reasons. Due to
the presence of impurities in even the highest quality solvents, it was
necessary to distill each of the above solvents before use. The use of
three solvents instead of two added a further burden to the laboratory
-37-
-------
0.4 -
0.3
O.D.
0.2
0.1
Start
CH3OH
LCA
LCO
mfi-
FIGURE 5 TYPICAL ALC CHROMATOGRAM OF
A DIESEL EXHAUST SAMPLE
-38-
-------
operation, and the volatile methylene chloride-hexane mixture compo-
sition could not be guaranteed over a long period of time. Our ob-
jectives for the whole instrument program have been to make each compo-
nent as simple for the operator and as reliable as possible.
An acceptable solvent must not absorb at 254 nm; must resolve
the atomatics from the oxygenates to yield only two fully resolved peaks
within a reasonable period of time (5 - 10 minutes), must be sufficiently
miscible to ensure column reactivation; must give a minimum background
contribution to the oxygenates peak; and if it contains impurities must
be possible to purify in a simple manner.
We have systematically looked at the 9 combinations of the
solvents:
Pentane Methanol
Hexane Ethanol
Cyclohexane Isopropanol
The results of our studies can be summarized as follows:
• Each of the three hydrocarbons, when used with methanol,
tends to resolve the sample into three peaks, and the
chromatograms were not reproducible.
• Ethanol used with either of the three hydrocarbons
satisfactorily resolves the aromatics from the oxygenates
as two peaks, but gives a higher background compared with
isopropanol.
• Isopropanol gives a two-peak resolution of the aromatics
and oxygenates with pentane and hexane, but the column
is not reactivated sufficiently with these combinations.
• Pentane and isopropanal give virtually no background.
However, due to its low boiling point, an erratic
change in solvent flow rate occurs due to vapor lock
as the pump temperature increases with use.
• Cyclohexane and isopropanol (both Matheson Coleman
and Bell, Spectroquality grade) appear to be the best
combination. Each is a relatively high boiling solvent
and presents minimum volatility problems. The residual
impurities in the cyclohexane can be conveniently re-
moved by filtration through an activated silica gel
column. Further, this solvent combination leads to a
positive and negative component in the background peak
(refractive index effects) which ultimately cancel each
other in the integrated signal. Typical results with
this solvent combination are shown in Figure 6.
-39-
-------
a. Diesel Exhaust
b. Background
f
IPOH
m£ of solvent
of solvent
FIGURE 6 ALCCHROMATOGRAM OF DIESEL EXHAUST WITH
THE CYCLOHEXANE-ISOPROPANOL SOLVENT SYSTEM
-40-
-------
C. Detector Response
It was necessary to determine the quantitative relationship of
the UV detector response for LCA and LCD exhaust fractions from a range
of diesel fuels in order to be able to properly calibrate the instrument
and provide a direct mass emission (microgram) output. Calibration stan-
dards were also needed for instrument adjustment and check-out.
1. Fuel Studies - LCA Response
The characteristics of the ALC response as a function of fuel
property have been explored for a range of fuels obtained for this pur-
pose. The fuels and their characteristics are.lsited in Table 18. Most
of the ALC data had been determined previously "*' and is on a peak height
basis. We have seen that for the LCA fraction, the height or area appear
to be equivalent ways of representing the sample due to the symmetric
nature of the chromatographic elution.
In Figure 7 the OD/yg LCA is seen to be a closely correlated
linear function of the OD/yJl of fuel response. Therefore, it will be a
simple matter to calibrate the ALC response with the injection of a solu-
tion of the fuel being used in the engine. The open circles are the kero-
sene and 50/50 mix results which were not used in the initial correlation.
From Figure 8 the fuel OD/y£ response also appears to be related
to the aromatics content on a semi-logarithmic basis.
2. Fuel Dependence of ALC LCO Response
The work above has shown that there is a substantial effect of
fuel composition on the LCA detector response in the ALC instrument and
that this effect must, therefore, be accounted for in determining the
mass (yg/£) of the LCA fraction. The influence of this composition ef-
fect on the LCO detector response was also studied.
Samples of exhaust from No. 1 and Midwest No. 2 fuel have been
collected for careful GC mass determination of the LCO fraction and measure-
ment of the ALC area/yg response. These data, together with the other
available data, are summarized in Table 19 and shown in Figure 9. Each
data point represents a separate exhaust sample.
Within the precision of the data, the LCO optical density (area)
response per yg from each type of fuel appears to be the same. For com-
parison, we have included the data obtained from our study of jet combus-
tor exhaust odor conducted for NASA*. In this case the fuel type included
isooctane, ASTM Al, JP-4 and JP-5. The average of all of the NASA response
data (1.08 area/yg) is the same as the diesel data (1.11 area/yg) indicating
* D. A. Kendall and P. L. Levins, Odor Intensity and Characterization of
Jet Exhaust and Chemical Analytical Measurements, final report to
National Aeronautics and Space Administration, March 1973, Report No.
NAoA CR—121159.
-41-
-------
Table 18
ALC
Fuel
Kerosene 61611
Kerosene 60064
Kerosene 61338
Kerosene 61835
No. 1 Diesel
No. 2 East Coast
No. 2 Midwest
No. 2 61836
No. 2 61339
No. 2 61610
No. 2 60063
East Coast Heating
oil
50/50 No. I/No. 2
Midwest
Kerosene
Soltrol
LCA Fuel
Sp. Gr.
0.812
0.808
0.820
0.811
0.832
0.826
0.852
0.849
0.863
0.852
0.837
0.854
_
0.818
-
Response Characteristics
% FIA
Aromatics
16.5
16.0
16.5
16.5
20.8
24.1
34.7
32.0
37.5
30.5
28.0
35.4
27.8
15.5
1.5
50%
Dist (°F)
405
414
421
415
436
482
506
486
508
504
472
498
-
4.8
—
ALC OD per
y£ fuel
1.19
1.36
1.46
1.81
2.08
5.6
11.8
9.2
12.9
11.8
12.7
19.3
6.5
1.5
0.13
Mg LCA
0.0089
0.0105
0.0108
0.0135
0.0120
0.030
0.040
0.034
0.040
0.046
0.054
0.064
0.028
0.012
—
(1) UV response determined using optical density
peak height technique.
-42-
-------
0.06
0.05
OD/M9 LCA
0.04
0.03
0.02
0.01
OD//ufi = -1.94 + 315 (OD/M9 LCA)
r2 =0.96
a -=1.28
2a = 2.6 (OD/jifi)
I
10
OD/yA Fuel
FIGURE 7 FUEL ALC RESPONSE
20
-43-
-------
40
30
matics 20
10
0.1
1.0
10
OD/yJl
FIGURE 8 RELATION OF ALC RESPONSE TO AROMATIC CONTENT
-44-
-------
Table 19
Dependence of ALC Detector Response :
GC Data'1'
Sample
Chrom 60
Chrom 62
Chrom 84
Chrom 85
Chrom 86
Chroir 88
Chrom 89
Chrom 90
Chrom 66
Chrom 67
Chrom 91
Chron 92
Chrom 93
Chrom 64
Chrom 65
NASA 27
NASA 36
NASA 46
NASA 47
NASA 48
NASA 49
Fuel
No. 1
Kero
No. 1
No. 1
No. 1
No. 1
No. 1
No. 1
No. 2
No. 2
No. 2
No. 2
No. 2
50/50
50/50
Iso
ASTM
Iso
JP-4
JP-4
JP-5
yg Reproducibilitv (%)
0.90
1.53
4.00
2.70
2.85
1.75
1.25
1.35
2.80
3.65
1.45
1.40
0.75
3.00
4.60
1.25
1.05
2.80
3.25
2.75
5.90
-
-
2
7
10
-
-
-
-
32
24
32
21
-
24
-
-
11
22
26
14
ALC Response
Area
1.14
2.26
2.92
3.24
2.34
1.26
1.35
1.20
3.08
5.48
1.65
1.69
1.05
3.12
5.10
1.98
1.21
2.58
2.76
2.20
4.43
Area/yg
1.27
1.48
0.73
1.20
0.82
0.72
1.08
0.89
1.10
1.50
1.14
1.21
1.40
1.04
1.11
1.58
1.15
0.92
0.85
0.80
0.75
(1) yg a'nd area per 5 yJl of sample solution. This is the
sample size used for the GC and ALC analyses.
-45-
-------
O D
<
CO
<
s
s
I
X
D
O
O
O
CO
I
co
UJ
cc
O
UJ
UJ
Q
8
O
O)
111
oc
D
CO
°i
ra
I
-46-
-------
further that there is no significant fuel composition effect.
We feel that much of our observed variation can be due to the
poor reproducibility of the GC mass data. The data show a range of re-
producibility of 7 - 32% in the GC data, as we have reported before.
The reproducibility of the ALC data is in the 2-5% range.
The ALC instrument has been designed to adjust for both LCA
and LCD fuel reponse differences, but a single value for the LCO re-
sponse has been used at this time.
3. Calibration Standards
Several reference chemicals have been studied for their ALC
UV detector response in order to select a standard to be used for cali-
bration of the instrument for LCA and LCO response. A list of the chem-
icals studied and their area/yg detector response is given in Table 20.
Compounds were chosen which were known to be related to the species ob-
served in the diesel exhaust odor fractions.
The considerations for selection of the standard for each ex-
haust fraction were stability, availability, purity, and similarity in
elution pattern in the ALC when compared with an authentic exhaust sam-
ple. With all of these factors considered, the preferred choices are:
LCA - Indan
LCO - 2-Hydroxy-4-methoxyacetophenone (HMA)
The response characteristics of these standards compared to
diesel exhaust samples are as follows:
Detector Response (area/ug)
Sample LCA LCO
No. 1 fuel exhaust 0.225 1.1
Midwest No. 2 fuel exhaust 0.52 1.1
Indan 0.185
A-Hydroxy-2-methoxyaceto-
phenone(HMA) " '
-47-
-------
Table 20
Reference Compound ALC Detector Response
Area/yg
Compound LCA LCO
0.03
0.03
Indan(1) 0.185
Tetralin 0.056
2-Indanone 0.64
6-Methoxy-l-tetralone 1.1
1,2-Dimethoxybenzene 0.08
o-Methoxybenzaldehyde 2.1
4-Hydroxycoumarin 0.10
6-Methylcoumarin 0.66
Coumarin 0.85
2-Coumaranone 0.03
3-Hydroxyacetophenone 2.03
(2)
2-Hydroxy-4-methoxyacetophenone 0.79
2-Methyl-l, 4-naphthoquinone 3.2
Piperonal 0.48
Methylcyclopentenolone 3.3
(1) Chemical Samples Company, Columbus, Ohio
(2) Aldrich Chemical Company, Inc., Milwaukee, Wisconsin
-48-
-------
Calibration solutions should be prepared as follows:
LCA (for No. 1 fuel)
Dissolve 6.1 yg/y£ (0.61 g/100 m£) of indan in cyclohexane.
Since the indan/LCA response is 0.185/0.225 = 0.82, then a
5 yH aliquot of the calibration solution is equivalent to
25 yg of LCA fraction from No. 1 fuel exhaust [5(6.1)(0.82) = 25],
LCO
Dissolve 0.84 yg/y£ (0.084 g/100 mfc) of 4-hydroxy-2-methoxy-
acetophenone (HMA) in cyclohexane. Sine the HMA/LCO response
is 0.79/1.1 = 0.72, then a 5 y£ aliquot of the calibration
solution is equivalent to 3yg of LCO [5(0.84) (0.72) = 3.0].
In summary, the instrument should be adjusted to give the following re-
sponse with the calibration solutions:
5 y£ of Calibration Solution
Standard Response
LCA(Indan) 25 yg
LCO(HMA) 5 yg
D. ALC Instrument Description
The basic configuration of the ALC instrument is shown schema-
tically in Figure 10. Front and rear photographs are shown in Figures
11 and 12. Protective covers for the electronic modules were removed
for the photograph. The relationship of the timing control logic to
operating components is represented in Figure 13.
The system was designed to use commercially available components
wherever possible. Main purchased components are the Milton Roy Mini
pump, Carle injector valve and actuator, LDC UV detector, and Chromatronix
septum injector. The remaining-hardware components are assembled with
Swagelok connectors. All parts in contact with solvents or sample are
stainless steel, Teflon or glass (reservoir level indicators). The
timing logic and analog computer are all solid state devices designed and
fabricated by ADL. The system is described functionally in this report.
Complete blueprints of all components are available on request.
The instrument is designed so that it can be used in a number of
modes — from a push button routine analyzer to a research analysis tool.
-49-
-------
z
oc
D
O
U
tc.
UJ
O
oc
I-
V)
u
<
cc
3
O
u.
-50-
-------
FIGURE 11 DIESEL ODOR ANALYSIS INSTRUMENT
-51-
-------
FIGURE 12 REAR VIEW OF DIESEL ODOR ANALYSIS INSTRUMENT
-52-
-------
O 3
P
< O
o
111
I
O
O
o
'81
en
c
oc
LJJ
a.
O
O
z
O
cc
(-
o
o
o
z
O >
-c 15
8>
< ^
LL
O
X
V)
111
DC
8
-------
Thus, it has control options that allow it to be operated completely
automatically after sample injection by pushing the "Start" button,
or each of the functional components (pump, valves, gain control, etc.)
may be operated manually. The analog signal processor is also designed
so that the instrument may be used to analyze exhaust where the compo-
sition of the fuel is unknown beyond distinguishing between a general
No. 1 or No. 2 fuel. Optionally, where a fuel of known composition is
available, the analyzer (LCA portion) may be calibrated specifically
with a solution of the fuel.
The basic calibration, control, and signal selection options
on the front panel (Figure 11) are:
Push Buttons
Start
System Reset
Integrator Clear
Timers
Delay after Start before start of integration (T^)
Alcohol inject and valve refill sequence (12)
Gain change from LCA to LCD (13)
End data cycle (recorder off)(T4A)
End column recondition time (pump off)(T4B)
Potentiometers
LCA and LCD Calibrate
LCD background correction (System Blank)
Exhaust sample size (£) for TIA calculation
Switches
On/Auto/Off options on all components
No. I/No. 2 fuel selector for LCA gain change
Recorder output - 2 channels, each with the option
of OD, yg, TIA
The basic operation is most easily explained for the automatic
cycle mode of operation. The timing logic is described schematically in
Figure 14. After injection of a sample, the "Start" button is pushed.
After a preselected time T-l, the integrator output is activated for
recording the area of the LCA peak (and, therefore, yg or TIA). The
T-l delay is allowed in this interval to minimize the effects of possible
system drift. At time T-2 the isopropanol is injected in order to elute
the LCO peak, the remaining series of operations (T2A, T2B, T2C) required
to reposition and fill the valve are also sequenced at this time. After
-54-
-------
P CO
u
CD .
CD
r
IT
o
N
f-
f
CD
2
61 S
W UJ
(_ ->
oc ?
o>.
—ICJiJ-
SUJ UJ
HOC
' 0
T
1
o:g
11
=•2
UJI-
QO
OO —
t-l-UJ
OCCC>
<
oc
Q
U
O
I
O
^
UJ
tc.
-55-
-------
the LCA peak has eluted (at T-3), the integrator gain is switched to
the proper gain control for the LCO fraction. After the LCO peak has
eluted (at T4A), the recorder and integration are stopped, but the pump
continues for an additional time (T4B) required to recondition the column
and prepare it for the next samples. The time^ are all determined by
injection of the LCA and LCO calibration standards. Normal complete
cycle time is 5 minutes.
1. Timing Circuit Description
Timing control is provided by a 60 Hz clock divided to appro-
priate frequencies. The front panel timers (digital codes) control the
selectable time options while the fixed time sequences are set on rear
panel timers.
The heavy duty AC loads (pump, valves, etc.) are driven by
solid state relays, or the associated manual switch. The DC loads
(analog circuit relays, display lamps) are controlled by transistor
drivers.
The display lamps are lit at various times during the operation
of the instrument to show the state of the system.
2. Analog Computer Signal Output
The analog computer circuit takes the basic OD output from the
UV detector and conditions the signal, based on all of the correlations
developed in the program, to provide output signals in terms of OD, yg
or TIA of the LCA and LCO fraction. The basic signal conditioning steps
involved can be represented in an oversimplified way in the Schematic
of Figure 15.
The first buffer amplifier is to convert the 0 - 0.64 full scale
response of the UV detector to a more convenient 0-1.0 full scale re-
sponse (although the response is still only linear to 0.60 OD). The
operational amplifier converts the OD response into yg of LCA or LCO
through the buffer amplifier by input of the detector response calibra-
tions determined as reported in earlier sections. Gain control is pro-
vided for the difference in LCA detector response for No. 1 and No. 2
fuel and also for any potential future difference observed for LCO, al-
though at the present time all LCO responses are taken as equal.
An LCO background offset (System Blank) correction is included
so that if a blank run in the ALC shows a signal (+ or -) in the final
integrated ye value for LCO, this may be null balanced to an effective
zero and, therefore, eliminated from all later analyses. In order to
maintain good signal to noise characteristics, the output from this
stage is in ug for total amount of sample injected. Thus, if a 0.5 I
-56-
-------
UV Detector
Output
Buffer Amplifier
Operational Amplifier
Optical Density, 0-1.0 F.S.
LCO Background
Correction
Buffer Amplifier
Gain for No. 1,
No. 2 Fuels
LCA,LCO
Log Module
LCA, 0-100jug F.S.
LCO, 0-10 jug F.S.
Sample Size.C
Buffer Amplifier
LCA, LCO
TIA, 0-3 F.S.
FIGURE 15 UV DETECTOR ANALOG COMPUTER SIGNAL PROCESSING
-57-
-------
aliquot of exhaust sample were injected, the observed yg value should
be multiplied by 2 to obtain
The log module converts the ug value into TIA units using the
correlation values discussed earlier. In this case the sample size can
be included and, therefore, the correct final TIA value is observed.
A typical result obtained from a 0.5 I sample of No. 2 fuel ex-
haust is shown in Figure 16. For this sample the LCA value is 2 x 13 =
26 ug/£, the LCO value is 2 x 2.6 = 5 ug/£, and the exhaust odor inten-
sity has a TIA value of 1.7.
3. Interpretation of Data
At the present time the data base is sufficiently accurate and
represents a wide enough range of emission conditions to
• accurately determine values (pg) of LCA
and LCO
• predict exhaust TIA from LCO values
There have not been sufficient odor studies on separate LCA and
LCO fractions to estimate their individual odor intensities. Eventually
the data base should be developed so that the individual odor intensities
could be predicted for both the LCA and LCO odor fractions. One could
then rank the relative odor intensities of the oily-kerosene and smoky-
burnt odor character groups.
In a complex mixture, the TIA value for the mixture is equal
to the intensity of the most intense odor note in the group. Thus,
with diesel exhaust, the TIA value for the exhaust would be the value
for LCA or LCO, whichever is greatest. In all of our work to date,
the LCO (smoky-burnt) odor group has always been the most intense.
Thus, we assume that the correlation for exahust TIA and LCO
(TIAExh =1.0+1.0 log LCO) is the same as for LCO odor intensity, but
this assumption should be verified.
A very limited amount of data from our tests suggest an approxi-
mate correlation for the oily-kerosene odor fraction of TIA~, = 0.4 + 0.7
log LCA. This correlation should be refined with a more extensive set of
data, comparable to that used for the total exhaust correlation.
-58-
-------
/jgLCA
TIA
Cyclohexane-lsopropanol
F = 1 mC/min
15 cm SS x 0.635 cm O.D.
Corasil II
0 r-
40 '-
4 -
a) Background
A ILL
3 -
2 -
b) 0.5 C of exhaust
No. 2 Fuel
1 -
O.D.
4 ' / 0 2
mC of solvent (and min)
FIGURE 16 ALC INSTRUMENT ANALYSIS OF DIESEL EXHAUST
-59-
-------
-------
VI. RECOMMENDATIONS
This diesel odor research program has been successfully com-
pleted. The chemical composition of diesel exhaust odor has been deter-
mined, and an instrumental method developed for its measurement. Now
that the initial needs have been satisfied, there is good reason for
making recommendations for future studies which will improve the odor
analysis capabilities and reduce the odor emissions from diesel engines.
Those areas of research which seem most appropriate for the next study
phase are discussed below:
1. ALC System Details
There are still several points related to the direct use of
the ALC instrument that should be studied. Sample collection details
and procedure should be defined quantitatively concerning sampling rates,
temperatures, volume measurement and sample trap geometries. The fuel
dependence of the area/yg detector response for the LCO fraction should
be confirmed with a larger data base. The TIA-concentration dependence
should be determined for the isolated LCA and LCO odor fractions.
2. User Evaluation of ALC
The ALC instrument should be used by several research labora-
tories to identify possible problems and to verify its utility and re-
liability.
3. Extended Data Base
Odor and instrumental measurements should continue to be col-
lected to cover as wide a range of conditions as possible in order to
improve the statistical accuracy of the ALC odor prediction and to
identify possible areas of departure from the trend identified thus far.
A wider range of fuels and engines and engine operating conditions should
be examined.
4. Diesel Exhaust Odor Measurement
Studies should be initiated to obtain the basic data to relate
measured values of odor intensity to the community response to diesel
odor. Such studies could include odor survey, dilution measurements,
observer differences in odor response, and odor character effect on ob-
ject ionability.
-61-
-------
5. Reducing Emissions
Appropriate research laboratories should begin to study means
of reducing diesel exhaust odor emissions. The ALC instrument response
should be used as the basis for measuring the odor intensity levels.
-62-
-------
VII. REFERENCES
1. Chemical Identification of the Odor Components in Diesel
Engine Exhaust, final report July 1969, CRC Project CAPE-
7-68, HEW Contract PH 22-68-20.
2. Chemical Identification of the Odor Components in Diesel
Engine Exhaust, final report June 1970, CRC Project CAPE-
7-68, HEW Contract No. CPA 22-69-63.
3. Chemical Identification of the Odor Components in Diesel
Engine Exhaust, final report June 1971, CRC Project CAPE-
7-68, EPA Contract No. EHSD 71-18.
4. Analysis of the Odorous Compounds in Diesel Engine Exhaust,
Final report June 1972, CRC Project CAPE-7-68, EPA Contract
No. 68-02-0087.
5. H. S. Ford, D. F. Merrion, and R. J. Hames, "Reducing
Hydrocarbons and Odor in Diesel Exhaust by Fuel Injector
Design", SAE Paper 700734 presented in Milwaukee, Wisconsin,
September, 1970
-63-
-------
-------
APPENDIX A
ODOR PROFILE MEASUREMENTS
1. DIESEL EXHAUST ODOR
The standard diesel exhaust sample has normally been obtained
from a 4-71 diesel engine operating at 33% load and 1800 rpm on No. 1
diesel fuel. The profile analysis of the standard sample diesel exhaust
was consistent from day to day, but the odor of diesel exhaust did show
some differences with variations in engine operation. Preliminary studies
indicate that within the normal procedural time interval between sampling
and examination, there are no detectable losses. Indeed, the odor appears
to persist with only slight change for over one hour. Total profile char-
acterization is consistent with the odor observed when traveling behind
a bus, which confirms our belief that the mode of engine operation pro-
vided a representative sample for analytical studies.
The description of the diesel exhaust odor in the static test
room with a dilution ratio of 600:1 can be described by three character
notes; oily, represented by technical grade hexadecane among other stan-
dards; burnt, which, although similar to a low dilution of propionaldehyde,
phenol, and cresol, is produced in fuels with partial oxidation at ele-
vated temperatures; and kerosene, which is the top odorous component of
the fuel and may be described as having sweet, sharp, sour, tarry,
solvent components. In addition to these odor characteristics which ap-
peared in the slight-to-moderate intensity range at this dilution, two
feeling sensations - nose irritation and eye irritation - were apparent.
As implied by the descriptive terminology used, some of the
odor characteristics are present in the fuel itself. The odor charac-
terization of a 150-y£ aliquot of fuel, which (by computation) is equiv-
alent to the amount of fuel burnt to produce the 21-liter (sample of ex-
haust, produces an odor in the test room at least as strong as the exhaust
odor. The dominant odor characteristic is kerosene, with the oily note
being less intense and the burnt aroma barely detectable. With the diesel
exhaust, the oily and burnt aromas are primary character notes and kerosene
a supplementary factor.
* Taken from the Final Report on CRC Project CAPE-7-68 and EPA Contract
No. 68-02-0087, June 1972, "Analysis of the Odorous Compounds in Diesel
Engine Exhaust."
** Detroit Diesel Allison Division of General Motors Corporation, 4-71
diesel engine-generator set; Model No. 4154N
A-l
-------
2. ODOR PROFILE METHODOLOGY
The Odor Profile Method of analysis has proven useful in
flavor and odor studies in a wide range of food and nonfood products.
The Profile Method, which originated at Arthur D. Little, Inc., 25
years ago, is a semiquantitative and qualitative description of the
odor sensation. The total odor sensation can be described by six
character notes. The method is qualitative in that there is verbal
description as to the odor quality(s) perceived. The order of appear-
ance of odor character notes indicates the other odor qualities pre-
sent as a function of time on a microsecond basis.
The intensity of each character note (as well as the Total
Intensity of Aroma, TIA) is rated on a four-point scale ranging from
threshold-)(, slight-1, moderate-2, to strong-3 intensities. It has
been our experience that for the odor intensity to increase by one unit
(e.g., from slight to moderate), a ten-fold increase in concentration
is required. The threshold intensity indicates that the character note
detected is just recognizable. The basic four point scale of threshold-
)( to strong-3 intensity can be expanded into a seven point scale with
experienced panelists by the use of one-half ratings. Thus, the full
scale of intensity rating of the odor strength is summarized below:
Numerical Rating Intensity
)( Threshold (recognition)
^2 Very slight
1 Slight
1^2 slight to moderate
2 Moderate
2% Moderate to strong
3 Strong
The presence of feeling sensations is indicated by a check
mark (/) without any effort to describe their intensity. Four trained
analysts form the odor profile panel. The sample to be analyzed is pre-
sented to the panelists in a standard manner. In this study, each of
the four panelists entered the odorized test chamber independently of
one another and sniffed the air three times. Each then recorded his
observations on the odor character notes perceived, their order of ap-
pearance, and their intensity. After the observations in the test room,
the panelists gathered to discuss their results. Reference was made to
odor standards to relate the various verbal descriptions used and to
develop common language in describing the odor quality. Reference odor
standards may be single chemical species or may refer to a mixture of
chemicals.
A-2
-------
The panel's results were then composited into an odor pro-
file that summarized the odor observations of the four panelists and
indicated the odor quality, the order of appearance of the characteristic
notes and their intensities.
3. STATIC ODOR TEST ROOM
The static odor test room consisted of an antechamber, an
odor chamber, and supporting equipment such as fans, ducts, activated
carbon, air intake, and air exhaust motors. The air is treated with
activated carbon (C-42 cannister from Dorex) and provided a low-odor
background diluting medium and was also used to flush odorized air from
the chamber and acclimate the four panel members to a low odor back-
ground. The odor chamber where the odor studies were carried out was
an aluminum-clad room with a volume of 12,600 £. Previous studies have
shown polished aluminum to be satisfactory for odor studies because it
has a low odor background. Fans in both the odor chamber and antechamber
ensured adequate mixing and assisted in flushing the test room with odor-
free air.
The sequence of events occurring prior to an odor observation
by the panel in the test room is listed below.
a. Odor-free air is used to flush out the antechamber
and odor chamber.
b. The door connecting the odor chamber and antechamber
is closed, thus sealing the odor chamber.
c. Diesel exhaust is injected into the odor chamber
through a sampling line by means of a swivel-jointed
sampling system. Most of the initial odor studies
were done using a 600/1 dilution of exhaust obtained
from a 21-liter sample. Three fans located in the
odor chamber circulate the diesel exhaust with the
diluting air to ensure proper mixing. Five minutes
after injection the fans are shut down, and the odor-
ized air in the odor chamber is allowed to come to
rest.
d. The four panel members then enter the antechamber
where they become acclimated to the low odor back-
ground air.
e. The panel then enters the odor chamber to make ob-
servations.
A-3
-------
f. The cycle is then repeated to prepare the odor
test room for the next observation. A 20-minute
flushing period has been found to be adequate for
removing odor from the test room.
4. DYNAMIC ODOR TEST CHAMBER
A dynamic odor test chamber has been constructed at ADL in
conjunction with other research progrms being conducted on the basic
characteristics of odor. This facility, shown schematically in Figure
A-l, provides an excellent means for efficiently conducting the dose/
response odor studies required for more precise determination of the
odor intensity of exhaust samples.
The facility has been constructed based on the same general
guidelines used for the static chamber. It consists of a carbon filter
to provide normally a 100 £/sec odor-free air flow which is controlled
by the exhaust fan. A remote controlled syringe drive introduces TOE,
LCA, or LCD exhaust samples which are then thoroughly mixed in an ADL
static mixer to insure homogeneous sample presentation in a laminar
flow through the test chamber. The chamber (approximately 2 x 2 x l.lm)
comfortably seats a single observer. Communication for reporting odor
intensity and character is achieved through a microphone/headset com-
bination.
During a test the odor analyst is not informed as to the
time or order in which samples are presented. The analyst is instructed
to report observations when perceived. A typical procedure would be as
follows:
1. Analyst enters chamber, adjusts microphone,
and adapts to odor-free background for three
minutes.
2. A sample is injected for a period of 30 sec.
at a predetermined concentration. Odor ob-
servations when perceived are reported verbally
where they are recorded by the technician operating
the facility from an adjacent room.
3. An odor-free background is presented for two
minutes.
4. Sample presentation is conducted at a new con-
centration.
5. The sequence is repeated for the number of points
required.
A-4
-------
GO
<
X
u
(/>
LLJ
I-
cc
o
a
o
o
<
z
a
tr
D
C3
3
A-5
-------
Each sample is examined by four panelists, and the average TIA value is
reported. Typical results for a TOE total exhaust extract would be:
Dose Concentration TIA
Ql/m3) CAverage of 4 member panel)
0.2 0.5
0.4 1.0
0.8 1.5
1.6 1.75
3.2 2.5
break
0.8 1.25
0.2 0.5
3.2 2.25
1.6 1.75
0.4 0.75
When these data are treated by the established semi-logarithmic relation-
ship of sensory response, a dose/response curve such as shown in Figure
A-2 is obtained. From the least squares correlation of these data, the
exhaust odor intensity can be described as
TIA = 1.50 + 1.50 log Cone.
2a = 0.4 TIA
The odor intensity at 1 £/m3, corresponding to a 1000/1 dilution is
mathematically represented by the constant in the equation above. It
is thus convenient to report the observed odor intensity in terms of
the constant and slope of the line.
A-6
-------
cc.
O
o
O
CO
t LU
a? _J
I w
1 CO
Z UJ
2 5
cc
I-
z
LU
o
o
h-
co
s
O
Ul
Q
UJ
CO
o
85
UJ
cc
UJ
tM
N
•H
CO
CM
A-7
-------
-------
APPENDIX B
CHEMICAL COMPOSITION OF THE ODOR COMPONENTS IN DIESEL EXHAUST
A detailed chemical description of the odorous compounds re-
sponsible for the characteristic diesel exhaust odor is given in the
final reports of the CAPE-7-68 project (refs. 2-4). The conclusions
of these studies are presented here for review purposes. In the sensory
evaluation of diesel exhaust odor, we have described the odor as having
two primary odor character groups, oily-kerosene and smoky-burnt, each
contributing about equally to the exhaust odor intensity. The compo-
sition of each of these groups is summarized below.
1. Oily-kerosene
The oily-kerosene odor character group is isolated from diesel
exhaust samples in the aromatic (LCA) fraction using the liquid chroma-
tographic separation procedure (ref. 4). The primary chemical species
identified in this fraction are summarized in Table B-l. Using a No. 1
diesel fuel, the quantitative composition of this fraction was found to
be typically
Species Weight
Alkyl benzenes 24
Indans/tetralins/indenes 14
Naphthalenes 57
Completion of the two-stage gas chromatography-odor-high reso-
lution mass spectrometry analysis of this fraction revealed that the main
odor contributor responsible for the kerosene-type odor was the series of
alkyl indans and tetralins. The alkyl benzenes were found to contribute
to the oily odor character assisted by the indenes. No specific odor cor-
relation was found with the naphthalenes, although they may contribute
to the overall odor perception through some form of synergism.
B-l
-------
M
-
O
ui
Z
UJ
CO
O
CC
UI
O
cc
cc
UJ
UJ
Lk
Z
g
I
UJ
CO
B-2
-------
Further studies showed that the composition of the aromatic
fraction was the same as that isolated from the fuel. The principal
source of this odor fraction is, therefore, believed to be in the aro-
matic fraction of the unburned fuel in the exhaust hydrocarbons.
2. Smoky-Burnt
The chemical species responsible for the smoky-burnt exhaust
odor are much more numerous and complex than was the case for the oily-
kerosene odor group. The primary species found to be associated with
the characteristic odor are listed in Table B-2.
Of the species identified, the greatest contribution to the
smoky-burnt odor character appears to be from the higher molecular weight
components and those with multi-functional substitution. Feel factors
(irritation, pungency) are frequently associated with the lower molecular
weight members of a particular chemical class.
Summarized, we find that:
• The smoky odor character is most consistently associated
with hydroxy and methoxy indanones with some contributions
from methyl and methoxy phenols.
• Burnt odors are associated with furans and alkyl benzal-
dehydes.
• The oxidized oily character is usually ascribed to
alkenones, dienones, hydroxy cyclocarbonyls, and
indanones.
• Irritation factors seem most frequently to be associated
with the lower molecular weight phenols. Some benzaldehydes
and methoxy benzenes may also contribute to this sensation.
• While some unsaturated aldehydes contribute to a portion
of the exhaust odor complex, the most abundant exhaust
aldehydes do not appear to contribute significantly.
• Neither sulfur nor nitrogen-containing species contribute
to the smoky-burnt odor complex. Although such species
were observed during portions of the analyses, none were
associated with exhaust odors.
The major source of the most significant odor contributors
in this group appears to be the aromatic portion of the fuel, although
some contribution is also seen from the paraffin portion. Confirmation
of this observation was obtained with the study of a paraffin fuel
(Soltrol 200) where the odor character was seen to change markedly —
the odor was not eliminated, however.
B-3
-------
oc
CO
§9
51
DC
o
o
O
o
O Q
111 UJ
N N
9 9
X X
o o
HI < IM
a H a
^ z z
O OC OC
233
CO CD CO
111
N
CO
UJ
_i
O
(-
OC
<
a.
UJ >
5 a
>s
^ H
O <
S uj
CO -I
UJ
N
CO
UJ
_J
o
p
OC
CM
0)
H
,0
§1
oc °9
co ^
OC O
O 2
Q CO
o
UJ
a.
H-
UJ
OC
D
H-
U
D
QC
CO
UJ
UJ
°S
So
2l
< Q
OC OC
D D
u_ u.
N
z
111
en
u.
O
N
Z
UJ
00
UJ
O
>
I
UJ
uj uj ' -}
S Z T
ii s
< 2 i
Q Q <
2? §
B-4
-------
APPENDIX C
SAMPLE COLLECTION PROCEDURES
This appendix describes the sample collection procedures using
the large preparative scale and analytical size Chromosorb 102 traps. The
preparative size traps provide sufficient sample from 500 I of exhaust for
odor, liquid chromatography, and mass spectrometric analysis. The analyti-
cal size traps provide sufficient sample for the new ALC instrumental anal-
ysis procedure.
1. PREPARATIVE SCALE
A. Summary
These traps are normally used to collect the odorous components
from 500 - 1000 £ of exhaust. The resulting sample isolated from the trap
with pentane solvent is sufficient for all of the odor and analytical pro-
cedures developed in conjunction with this program.
B. Supplies
Stainless steel collection trap as shown in Figure C-l. (Draw-
ing is to scale, 1/1). Chromosorb is contained between the two 100 mesh
screens held in slight compression by the coil spring.
Glass funnel - 10 cm diameter with stem dimensions the same
as the trap (k in. O.D.) so that it may be close coupled- with Teflon
tubing.
n-Pentane (Phillips Petroleum, Chromatoquality) - distilled
at atmospheric pressure, B.P. 36°C.
Methanol (Fisher Scientific, Spectroquality).
Chromosorb 102 (Johns-Manville, 60/80 mesh).
C. Procedure
1. Chromosorb Preparation
The adsorbent is prewashed as follows: Plug the end of a
large 10 cm O.D. glass column with silanized glass wool and transfer in-
to it 200 g of Chromosorb 102. The Chromosorb is washed with methanol
C-l
-------
Bushing
Cap
Teflon
washer
0.635 cm (k in) OD
Teflon washer
cap
2.22 cm (7/8 in) OD Body
100 mesh S.S. screens
coil spring
Scale = 1/1
Figure C-l Preparative Scale Chromosorb 102 Diesel
Exhaust Sample
C-2
-------
until all traces of LCA and LCO are removed as determined by injection
of 10 uJ, of the eluate into the ALC instrument. About 1 I of methanol
is normally required for 200 g of Chromosorb. The column is allowed to
stand overnight and then washed with 500 - 800 mil of pentane. The pen-
tane is used to elute the methanol and facilitate drying.
2. Trap Preparation
Ten grams of Chromosorb 102 is weighed out and packed in the
stainless steel trap "between the Wire screens. Glass wool may also be
required if the screen mesh is too large. The direction of exhaust flow
is marked for later use in sample elution.
3. Exhaust Collection
The trap is connected to the exit side of a contamination free
pump (Metal Bellows MB-155) with Swagelok fittings. Exhaust (hot, fiber
glass filter filtered) is then passed through the trap at about 10 il/min
until the desired 500 - 1000 liters have been collected. After collec-
tion, the fastenings are removed and the trap capped with Swagelok plugs.
The sampling train is normally configured as indicated.
Exhaust
Heated
Particulate
Filter
Metal
Bellows
Pump
Chromosorb
102
Trap
Integrating
Volume
Meter
Rate
Meter
The temperature requirements for sampling are'not-known in de-
tail, although the normal hydrocarbon, analysis guidelines,should apply
in f'his case. Our particulate filter has normally,been .maiintained at ,
the exhaust temperature of 400°F. The Metal Bellows pump (1 c£m Mod,el
155) is not heated beyond its normal self heating from the work of ;the,
bellows.
L 4. ~ Sample.' Isolation . . , . _ „,
The trap is fastened upright on a ring. st;and .-^hoye a. gxadu^ted
receiver wit'h^the trap inlet side facing down, to the rec,e4-ver.j » ;TheKr
funnel is' then attached to the trap exit end withTeflon,, tubing,,'.'.'....; '..,
; A separatory funnel or solvent reservoir is positioned jrWS«tr
above the funnel and filled with about 10 mil ofJ'pentaneAXThe -solvent,,
is allowed to slowly percolate through the trap. Flow should be about
C-3
-------
0.5 m£/min. The solvent must be added to what had been the exit end
of the trap so that the trapped odor species may be eluted as easily
as possible. If the solvent flow is reversed, the trapped species
chromatograph on the bed packing with resultant lower recovery.
The first 10 mi of pentane effluent is collected and contains
all of the total organic extract (TOE).
2. ANALYTICAL SCALE
A. Summary
The analytical size traps are capable of collecting the odor-
ants from 10 - 50 I of exhaust for subsequent ALC instrumental analysis.
The basic method is the same as described above, differing only in de-
tail of scale.
B. Supplies
Two different configurations of analytical traps have been
used as shown in Figure C-2. Photographs of the preparative and ana-
lytical traps are shown in Figure C-3. The AC (Analytical Cjhromosorb)
traps were developed first. Using the Metal Bellows 155 pump, the nor-
mal sampling rate is about 2 £/min with these traps. The newer ACS
(ACShort) traps contain the same amount of Chromosorb 102, but the larger
cross-section normally allows sampling at a rate of 10 £/min.
Cyclohexane (Matheson Coleman and Bell Spectroquality) - the
solvent is filtered through 180 g of activated silica gel, Grade 950,
60 - 200 mesh (Fisher Scientific) packed in a 50 cm x 3 cm OD glass
column. If the column is kept wet with cyclohexane, up to 3 gallons
may be purified before the silica gel is deactivated.
C. Procedure '
Chromosorb 102 is prepared as described previously. One gram
of Chromosorb is packed into either the AC or ACS traps using silanized
glass wool plugs at each end to retain the powder. Exhaust samples are
collected as for the Preparative scale sample.
Samples are isolated from the traps for analysis using the fil-
tered cyclohexane and the same general procedure described above. Again,
one must be careful to elute the sample in the reverse direction from
which it was collected to avoid chromatographic losses on the Chromosorb.
The eluting solvent flow rate should be held to about 0.5 m£/min and may
require flow restriction on the ACS traps to achieve this rate. We have
successfully used hypodermic syringe needles adapted to the Swagelok
fitting for this purpose. The TOE sample is contained in the first 1 m£
of eluting solvent.
C-4
-------
x~ § PLUG
I •'_ i"
•5 8
REDUCING UNION
SILANIZED
GLASS WOOL
CHROMOSORB 102
TUBIKJG 5" LONG
.
8 8
REDUCING UNION
SILANIZED
GLASS WOOL
CHROMOSORB 102
TUBING 2" LONG
|"00 « .030 WALL
w^
b. ACS TRAP
a. AC TRAP
FIGURE C-2 ANALYTICAL SCALE CHROMOSORB 102
DIESEL EXHAUST SAMPLERS
C-5
-------
FIGURE C3 PHOTOGRAPH OF DIESEL EXHAUST ODOR
SAMPLE COLLECTION TRAPS
C-6
-------
BIBLIOGRAPHIC DATA !• Report No. 2.
SHEET EPA-650/2-73-025
4. I ii k and Subtitle
Application of Odor Technology to Mobile Source Emission
Instrumentation
7. Author/s)
P. L. Levins and D. A. Kendall
9. !><_ rlormint Organization Name and Address
Arthur D. Little, Inc.
15 Acorn Park
Cambridge, Massachusetts 02140
12. Sponsoring Organization \arne .ind Address
Environmental Protection Agency Research Triangle Park
Office of Research and Monitoring North Carollna 27711
Chemistry and Physics Laboratory
National Environmental Research Center
3. Recipient's Accession \ .
1
5. Report Date '
September 1973 J
6.
8. Performing Organization Rep'.
No.
10. Project 'Task/Work Unit V.
1A1010 26 ACV /7
11. Contract/ Grant No.
68-02-0561
13. Type of Report & Period
Covered
' Final June 1972-8/73 j
14.
15. Supplememar \ Notes
Jointly sponsored by the Coordinating Research Council (APRAC) Comm CAPE-7-68
16. Abstract-, ^n instrumental method has been developed which measures the odorous species in
diesel exhaust. Liquid chromatography is used to separate the exhaust sample collect-
ed on a solid substrate. The fractions containing the odor are sensed using a uv detec-
:or which does not respond to many of the species which are not odorous. A correlation
Is shown between odor observed by this method, and the odor observed by subjective panel;
^ wide range of diesel emissions was used to construct the correlation which may be
expressed as: TIA = a + b log LCO where TIA is the total intensity of aroma observed by
the panel on a dose response basis and LCO is the "smoky-burnt" or oxygenated fraction
)bserved on the liquid chromatograph. a and b are constants which depend on volume colle
:ed and instrument response. The instrument and methodology is described to perform odor
neasurements and conditions described where both a and b are 1.0. Collection and analys:
ire described as are the basic methodology to determine TIA using odor panels and the
summary of chemical composition of diesel exhaust.
17. Ke\
and Document Anai>sis. 17a. I >r s
inputs _ odor, Diesel engine, exhaust composition
analysis, odor measurement, odor instrument
)7b. identifiers Open-Ended Terms
17c. COSATI Field/Group
13B
18. Availability Statement
Jnlimited
19. Security ria'>s TThis
Report i
UNCLASSIFIED
20. Security Class 'This
Page
UNCLASSIFIED
21. No. of Pages
87
22. Price
~ORM NTIS-35 'REV. 3-72)
USCOMM-DC I4952-P72
C-7
-------
------- |