National Enforcement Investigations Center
EPA-330/1-81-001
AUTOMATED ATOMIC ABSORPTION
DETERMINATION OF LEAD IN GASOLINE
February 1981
O.S. Environmental Protection Agency
r i
Office of Enforcement

-------
UNITED STATES ENVIRONMENTAL PROTECTION AGENCY
OFFICE OF ENFORCEMENT
EPA-330/l-81-001
AUTOMATED ATOMIC ABSORPTION
DETERMINATION OF LEAD IN GASOLINE
February 1981
J. H. Lowry
T. J. Meszaros
L. Conlon
NATIONAL ENFORCEMENT INVESTIGATIONS CENTER -

-------
INTRODUCTION
The determination of lead in gasoline by atomic absorption spectro-
metry has been adopted by the American Society for Testing Materials [1]
and the Environmental Protection Agency [2] as the standard method of
analysis. The method consists of the manual preparation of an in-situ
reaction of the alkyl lead compounds in gasoline with iodine, stabilization
of the alkyl lead iodide complexes with tricapryl methyl ammonium chloride
(Aliquot 336), ten-fold dilution with methyl isobutyl ketone (MIBK) and
measurement by atomic absorption spectrometry with an air-acetylene flame.
The iodine reaction eliminates the problem of variations in response due to
different alkyl lead compounds, Kashiki et al. [3]. The dilution compen-
sates for severe non-atomic absorption, scatter from unburned carbon, and
minimizes matrix effects, Lukasiewiez et al. [4].
The Environmental Protection Agency has initiated a yearly nationwide
survey to determine the extent of pollution control device tampering on
automobiles. Another purpose of these surveys is to determine if fuel
switching has occured, i.e., leaded gasoline is being used in automobiles
designed for use of unleaded gasoline (<0.05 g Pb/gal). Consequently,
these surveys require the analysis of numerous gasoline samples' lead con-
tent within our laboratory. This demand necessitates accurate, rapid
analyses by use of an automated method.
Heistand et al. [5] automated a nitric acid extraction atomic absorp-
tion method for the analysis of lead in gasoline. They stated that the
standard method could not be automated because pump tubing deteriorates

-------
2
rapidly in the presence of MIBK. The data we generate must be legally de-
fensible and must be comparable with data gathered by the manual standard
method. Hence, Heistand's automated method was deemed inappropriate for
our application.
The automation of the standard method is discussed below. The incom-
patability of the MIBK with the pump tubing was initially circumvented by
the use of solvent displacement flasks and later by use of constant flow
syringe pumps. Data showing equivalence of the automated and manual proce-
dures and precision and accuracy data gathered over a 4 month period during
the analysis of about 1,500 samples are presented. The effect of holding
times, container types and storage conditions on the lead content of gaso-
line samples was also studied. Findings indicate a definite need to speci-
fy these as requisites in the standard method.
EXPERIMENTAL
Apparatus
A Technicon Auto Analyzer Sampler and a Pump III were used for the
automated system. Standard heating block coils (#157-0225) were used for
mixing coils because of the need for good mixing with the high-flow rate.
A Perkin-Elmer Model 403 atomic absorption spectrophotometer and a strip
chart recorder were the detection system used for the manual and automated
procedures.

-------
3
The solvent displacement flasks are Erlenmeyer flasks fitted with sil-
icone rubber stoppers and glass tubing. Later, Model 220 Sage constant-
flow syringe pumps fitted with 20 m£ Teflon-coated syringes were utilized
for the addition of MIBK. The solvents are transported through Teflon tub-
ing fitted to the glass with heat shrinkable Teflon tubing. The glassware
employed in the manifold is interconnected with polyethylene tubing because
tygon tubing dissolved in the presence of MIBK. Solvaflex tubing was found
to be compatible with the gasoline. However, the iodine reagent solvaflex
pump tubing had to be changed daily.
Reagents
Working standards of lead alkyls in reference fuel (U.S., EPA, RTP,
N.C.) were utilized in both procedures. For the manual procedures the io-
dine solution (Fisher Sci. Co., Fairlawn, NJ) was 3% w/v in toluene (Bur-
dick & Jackson, Muskegon, MI). The automated procedure iodine solution was
0.24% w/v in toluene. In both procedures the Aliquat 336 (Aldrich Chemical
Co., Miswaukee, WI), solution was 0.88% in MIBK (Burdick & Jackson, Muske-
gon, MI). Certified unleaded gasoline was obtained from Phillips Chemical
Company, Borger, TX.
PROCEDURE
Manual
The procedure published in the Federal Register [2] was followed ex-
cept that alkyl lead compounds in isoctane standards were used instead of
lead chloride standards.

-------
4
Automated
The flow diagram of the automated system is illustrated in Figure 1.
A sampling rate of 30/hr with a 2:1 sample to wash ratio provides enough
peak resolution to establish a baseline at concentrations less than 0.05 g
Pb/gal. The procedure screened all samples at a 30/hr sampling rate. Any
samples with a lead content greater than 0.05 g/gal were rerun at a
sampling rate of 20/hr with a 3:1 sample to wash ratio. The wash solution
was certified unleaded gasoline. The sample was diluted and mixed in the
first mixing coil with MIBK displaced from a 2-liter flask with distilled
water. The iodine reagent (0.24% w/v) was then reacted with the air
segmented stream in the second mixing coil. At the flow rates given in
Figure 1, the reaction time of the iodine before the addition of the
Aliquot 336 was a little over 1 minute. The Aliquot 336 solution was
introduced into the system by means of displacement from a 500 ml flask
using distilled water. The air segmented stream was then debubbled by
reverse displacement and the products of the reaction were pumped into the
atomic absorption spectrophotometer. The operating conditions of the
Atomic Absorption Spectrophotometer were as follows: wavelength, 283.3 nm;
acetylene flow, 20 m£/min; airflow, 65 m£/min; nebulizer flow, 5.2 m£/min.
Sample Storage Study
Containers made of polyethylene, tin with soldered seams, and tin with
pressed seams were used in a sample storage study. In addition, tin con-
tainers with pressed seams were employed with and without a tin cap insert.

-------
5
Seven aliquots of the composite sample were analyzed to determine the zero
day lead content. Forty-eight 50-mil aliquots were transferred on the same
day to the individual sample containers. At intervals of 1, 2, 4 and 23
weeks, three samples of each container type stored at 4°C or at ambient
temperature (six samples/container type/interval) were allowed to come to
room temperature and then analyzed with the automated system. Subsequent-
ly, a similar study was performed with glass containers utilizing a 2-gal-
lon composite test sample.
RESULTS AND DISCUSSION
Manual and Automated Comparison
The additions of the individual reagents of the automated system were
designed to match as close as possible those of the manual method. Table 1
shows the volume of each reagent required by the manual method to the vol-
umes of reagents utilized in 1 minute by the automated method. Iodine at a
concentration of 3% w/v caused the pump tubing to harden very quickly. The
iodine was diluted and its pump flow rate was increased to result in an
equal molar concentration addition. The percent total volume and molar
concentration of each reagent is very closely matched in both methods.
The comparability between two methods of analysis is usually measured
by defining the sensitivity, precision, and accuracy of each method. Four-
point calibration curves were prepared over the concentration range of
0.010 g/gal to 0.110 g/gal for both methods. A least squares fit of the
calibration data for the manual and automated methods resulted in slopes of

-------
6
9.8 and 9.1 Abs. Units/g Pb/gal and intercepts of -0.016 and -0.012 re-
spectively. This demonstrates that both systems have about the same sen-
sitivity.
The percent relative standard deviation (% RSD) of replicate analysis
of a sample is a measure of the precision of the method. Table II shows
the results of the replicate analysis of three samples by the automated
method and the replicate analysis of one sample by the manual method. A
comparison of the % RSD of the respective methods indicates that the pre-
cision is very similar. Another measure of precision can be obtained by
the absolute difference of duplicate analysis. Four samples analyzed in
duplicate by both methods resulted in the data presented in Table II. The
average difference for duplicate analysis by both methods is less than
0.005 g/gal which is the maximum acceptable difference allowed by the
standard method. In consideration of these measures of precision, the pre-
cisions of both methods are about the same.
The accuracy of both methods was assessed by measuring the lead con-
tent of three NBS certified gasoline standards. The results of these
analyses are given in Table II. The manual analysis average deviation was
slightly biased high, 2.4% ± 7.2%, while the automated analyses are biased
low, -3.7% ± 1.8%. Absolute deviations would indicate that since the auto-
mated analyses fell within 5% of the true value while the manual analyses
were within 10%, the accuracy of the automated method is slightly better
than the manual method.

-------
7
Seventeen unleaded gasoline samples were analyzed by both methods.
The results of these analyses are given in Table III. A statistical stu-
dent T method comparison test indicates there is no statistical difference
between the results of the two methods.
There are a number of practical considerations that favor the use of
the automated analysis in our laboratory situation. When a workload of 20
samples is on hand, the automated analysis results in a 300% savings in the
cost of labor and reagents over manual analysis. In 1-man day, at least
100 sample analyses can be performed by the automated analysis while, in
consideration of glassware cleaning, sample preparation and analysis, only
about 25 sample analyses can be performed by the manual analysis. The ease
with which quality control data can be gathered with the automated analysis
offers an advantage that is of uppermost importance in producing legally
defensible analyses. An additional advantage of the automated system is
that since the entire system is closed, MIBK vapors are cut substantially
in the laboratory.
Survey Sample Analysis
Our in-house use of the automated procedure places a heavy emphasis on
quality control. This procedure requires first a check of the slope of the
calibration curve. All calibration curves used for sample analyses agreed
to within 10% of the slope stated earlier. Every tenth sample was analyzed
in duplicate. At least one NBS reference standard and at least one blind
RTP reference standard was analyzed during an analysis run. All samples
with a peak height greater than the 0.05 g/gal standard were rerun and

-------
8
spiked with known quantities of alkyl lead compounds. Samples were diluted
if necessary with unleaded gasoline so that the resulting diluted value, as
well as the diluted spike sample valve, fell on the calibration curve.
The quality control results of the analysis of 1,491 samples are sum-
marized in Table IV. The average difference of the duplicate analyses very
closely approximate zero which would be expected statistically. Within a
95% confidence interval all duplicate analyses performed with the automated
system would differ by less than 0.0046 g/gal which is within the accept-
able limit of 0.005 g/gal difference established in the standard method.
The accuracy of the method evaluated over an extensive period is quite
good. Within a 95% confidence interval, values reported over 0.05 g/gal
are within 10% of the true value. Values reported below 0.05 g/gal are
within 15% of the true value. As indicated by the spiking data, no sub-
stantial matrix effects were encountered in the analyses.
The percentage of unleaded gasoline-designed automobiles that switched
to leaded gasoline are summarized by state for the 1979 Eight State Survey
in Table V. The average fuel switching percentage was 9.3%. The differ-
ence between the Vermont I and II studies is that the Vermont II study pop-
ulation included a higher percentage of automobiles owned by rural persons.
Sample Storage Study
Polyethylene containers, tin containers with a lead solder seam, tin
containers with a pressed seam, tin containers with pressed seams with a
stainless steel insert, glass containers with linear polyethylene liners,

-------
9
and glass containers with Teflon liners were evaluated as to their effects
on the lead content of composite gasoline samples. Containers were stored
at 4°C and ambient temperature and analyzed at various periods of time.
Table VI summarizes the results of this investigation. The values reported
are an average of the analysis of three individual samples. The initial
lead content of the composition gasoline samples were 0.056 ± 0.014 g
Pb/gal for all the containers except glass, and 0.058 ± 0.006 g Pb/gal for
the glass containers. Initial values are based on seven analyses. At the
end of 1 and 2 weeks, the lead concentrations remained within the initial
value ranges. The polyethylene sample container stored at ambient tempera-
tures start to show a concentrating effect at the fourth week. The worst
case is the polyethylene container stored at ambient temperatures while the
best is the glass container. The concentrating effect is the loss of the
lighter weight gasoline fraction. Emission of these vapors is readily de-
tected by the odor emitted from the polyethylene containers.
The results of this study illustrate vividly the need for storage time
and types of containers to be specified in the standard method. This is
important to assure legally defensible analyses. In consideration of the
results presented in Table VI, all samples should be collected in glass
containers and analyzed within 4 weeks. Within this time frame, little
difference is observed between storage at 4°C and ambient temperatures.

-------
References
1.	"Am. Soc. for Testing Materials", Part 17, 1973. Method D 3237,
ASTM, 1916 Race Street, Philadelphia, PA 19103.
2.	Federal Register, 15449, July 28, 1974.
3.	M. Kashiki, S. Yamazoe, S. Ohitna, Anal. Chem Acta, 53, 95 (1971).
4.	Lukasiewiez, R. J., Berens, P. H., Buell, B. E. , Anal, Chem., 47,
1045 (1975).
5.	Heistand, R. N. , Shaner, W. C. , Jr., Atomic Absorption Newsletter,
13, 65 (1974).

-------
List of Figures
1. Flow Diagram for the Automated System (*: Solvaflex Pump Tubing,
**: Technicon part no. 157-0225.
List of Tables
I. Comparison of Reagent Usage by the Manual and Automated Methods
II. Precision and Accuracy Data for the Manual and Automated Methods
III. Statistical Comparison of Actual Sample Analyses by the Manual
and Automated Methods
IV. Precision and Accuracy Data for the Automated Method Gathered
During the Survey
V. Percentages of Automobiles by State that Switched to Leaded
Gasoline (0.050 g Pb/gal)
VI. The Effect of Time on the Lead Content of Gasoline Samples with
Temperature and Container Type

-------
(1.00)
AIR
DEBUBBLER
(2.90)
34T
TO AAS
(0.32*)
TO WASTE
(1.06*)
SAMPLE LINE
(0.16*)
DEBUBBLER
IODINE REAGENT
ALIQUAT 336
DISPLACEMENT WA 7ER (0-80)
(3.40)
TO WASTE
TO PUMP
34T
34T
MIBK
DISPLACEMENT WATER (2-90)
FINAL DEBUBBLER
DISPLACEMENT WATER (2.90)

-------
Reagent Proportions
Manual
Reagent	Automated
mL Used	% Total Volume	mL Used* % Total Volume
39 9	79 8	Ml BK	5 80	77 3
5 0	10 0	Std or Sample	0 74	9 9
0 1	0 2	I? /Toluene	0 16	21*
5 0	10 0	1% Aliquot 336/MIBK	0 80	10 7
50 0	100 0 Total Volume	7 50	100 0
Concentration of I?/Toluene solutions are 3% for manual method, 0 24^o for
automated method
*ML utilized in one minute.

-------
PRECISION
Replicate Analysis
No. of
Analyses
5
5
5
5
Method
Manual
Automated
Automated
Automated
Concentration
g Pb/gal
0.054
0.010
0.048
0.085
% RSD
3.6
4.2
3.5
3.3
Duplicate Analysis
Automated
Manual
Sample No.
Avg.
Difference
Avq.
Difference
1
0.031
0.003
0.033
0.001
2
0.043
0.000
0.045
0.002
3
0.012
0.004
0.015
0.001
4
0.101
0.003
0.098
0.008


= 0.0025
V
0.0030

sa
= 0.0017
CO
CL
11
0.0034

ACCURACY



NBS Reference Standards



Automated
Manual
NBS Value
Value
%
Value
%
g Pb/qal
g Pb/gal
Deviation
q Pb/qal
Deviation
0.0322
0.0307
-4.7
0.0350
+8.7
0.0519
0.0494
-4.8
0.0539
+3.9
0.0725
0.0713
-1.7
0.0685
-5.5

-------
iple
l
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
Found Value g/gal
Auto-Manual
Manual	Automated	Difference
0 032	0 032	0 000
0 044	0 043	-0 001
0 014	0014	0000
0 094	0 099	+0 005
0 035	0 034	-0 001
0 053	0 052	-0 001
0 012	0 007	-0 005
0 032	0 032	0 000
0 014	1 010	-0 004
0 086	0 092	+0 006
0 055	0 058	+0 003
0 032	0 030	-0 002
0 012	0 007	-0 005
0 050	0 055	+0 005
0 074	0 073	-0 001
0 010	0 005	-0 005
0 089	0_097	+0 008
x	= 0 000 12
difference
difference
= 0 004
n	= 17
t	=0123
For n-1 = 16 , t = 1.746
05

-------
Precision
Duplicate Analyses
n	Average Difference, g Pb/gal.
156	0 00011
Accuracy
NBS Reference Stds
n	Concentration,	g Pb/gal Avg % Deviation
21	0.0322	34
36	0.0519	0 7
20	0.0725	oj
"Blind" Reference Stds.
n	Average Difference	Std. Deviation
23	-0 0009	0 004
Spiked Samples
n	Average % Recovery	Std. Deviation
108	101	56
Samples Analyzed 1491
Std Deviation
0 0023
Std Deviation
6 4
4 8
4.9

-------
State
Peicent
Tennessee
9.7
Delaware
19
Minnesota
72
Vermont 1
152
New Jersey
1.6
Texas
104
Vermont II
29.1
Virginia
60
Arizona
22
Average
93

-------
TABLE VI
The Effect of Time on the Lead Content of Gasoline Samples with Temperature and Container Type
CONTAINER TYPE
Time
Period	Polyethylene	Soldered Seam	Pressed Seam	Pressed Seam w/Insert	Glass w/LPE Liner	Glass w/Teflon Liner
(wks)	Refrigerated/Ambient	Refrigerated/Ambient	Refrigerated/Ambient	Refrigerated/Ambient	Refrigerated/Ambient	Refrigerated/Ambient
1	0.057±0.009/0.05910.002	0.05110.010/0.06310.002	0.05410.007/0.05910.002	0.060±0.002/0.061±O.006	0.059±0.003/0.05810.000	0.05910.006/0.057±0.006
2	0.05810.012/0.06510.007	0.06l±0.008/0.063±0.011	0.057±0.006/0.065+0.009	0.060±0.007/0.06310.006	0.05810.012/0.05610.018	0.057+0.012/0.05810.012
4	0.061+0.006/0.07610.007	0.058+0.012/0.06910.005	0.05310.003/0.06010.005	0.05510.003/0.06510.006	0.05810.006/0.059+0.012	0.058+0.003/0.05710.003
23	0.100+0.007/0.15610.003	0.08710.009/0.09310.017	0.07310.005/0.07810.020	0.09210.022/0.08410.002	0.06110.002/0.06110.002	0.05710.003/0.05710.004
Values are averages for three samples 1 three standard deviations; g Pb/gal.
Initial Lead Content: 0.05610.014 g Pb/gal. for all but glass containers which was 0.05810.006 g Pb/gal.

-------