According to the European Committee for Standardization
(EN), both ultraviolet fluorescence (ISO 208461, sulfur by UVF) and wavelength
dispersive x-ray fluorescence (ISO 208842, sulfur by WDXRF) are applicable
test methods for fatty acid methyl ester (FAME), also known as biodiesel (EN
142143), and both are listed as
referee methods. However, the situation is different at ASTM International
(formerly the American Society for Testing and Materials).
According to ASTM D6751 “Standard specification
for biodiesel fuel blend stock (B100) for middle distillate fuels,”4 several options exist for
testing sulfur in biodiesel. ASTM D5453 (sulfur by UVF) is listed as the
referee method, but D7039 (sulfur by MWDXRF) may also be used. Additionally, the
D6751 specification also lists other sulfur test methods, but states that these
methods—D2622, in
particular—may give
falsely high results due to the presence of oxygen in biodiesel (typically 10
wt%–12 wt% oxygen).
This is a curious statement, as Section 6.2 of ASTM
D2622 “Standard test method for sulfur in petroleum products by wavelength dispersive
X-ray fluorescence spectrometry,”5 states the opposite:
“6.2 Fuels containing
large quantities of FAME, ethanol, or methanol (see TABLE 1) have a high oxygen content
leading to significant absorption of sulfur Kα radiation and low sulfur
results. Such fuels can, however, be analyzed using this test method provided
either that correction factors are applied to the results (when calibrating with
white oils) or that the calibration standards are prepared to match the matrix
of the sample. …”
* The concentrations of ethanol and methanol were calculated assuming
a theoretical mixture of hydrocarbons and di-butyl sulfide to which ethanol (or
methanol) was added until the sum of the mass coefficients times mass fractions
increased by 5%. In other words, the amount of ethanol (or methanol) that
caused a negative 5% error in the sulfur measurement was calculated. This
information is included in TABLE 1 to inform those who wish to use Test Method D2622 to
determine sulfur in FAME blends (biodiesel), gasohol, M-85 or M-100 of the
nature of the error involved.
How then has this discrepancy between the D6751 specification and
D2622 occurred?
D6751 RR:D02-1480
research report setup. Turning
back to the D6751 B100 specification, Appendix X1 for “Significance
of properties specified for biodiesel fuel,” Section X1.5.1 on Sulfur has a
Note X1.1 that states:
“NOTE X1.1—Test Method D5453 should
be used with biodiesel. Use of other test methods may provide falsely high
results when analyzing B100 with extremely low sulfur levels (less than 5 ppm).
Biodiesel sulfur analysis from RR:D02-1480, Biodiesel Fuel Cetane Number
Testing Program, January-April 1999, using Test Method D2622 yielded falsely
high results due to the presence of the oxygen in the biodiesel. Sulfur results
using Test Method D2622 were more accurate with B20 than with B100 due to the
lower oxygen content of B20. Potential improvements to Test Method D2622 may
provide more accurate values in the future.”
In 1999, an interlaboratory study (ILS) was initiated
by the Biodiesel Fuel Cetane Number Testing Program sponsored by ASTM
subcommittee D02.01 on combustion characteristics. For this study, a diesel
sample (sample ID 902) and three biodiesel samples (903B, 904B, 905B) were used
both as test samples and blend stock for additional B20 biodiesel blends.
Sample compositions were:
Multiple sample properties besides cetane
number were tested, including sulfur. Detailed directions for cetane testing
were provided to participants, but directions for the rest of the fuel
inspection tests (including sulfur) were lacking.
Also included in the study directions were expected
test results, or “inspection data,” for all sample properties, excluding cetane
number. It was not reported who conducted the initial inspection data testing,
or what ASTM methods were used in that testing. Note: It is unusual to provide expected
test results to study participants before the testing occurs. See TABLE 2 for a summary
of the sulfur data.
D6751
RR:D02-1480 research report results. TABLE 2 lists the sample ID, sample matrix,
number of participating laboratories, the sulfur inspection values (values
given to the participants prior to testing), the ILS results (labeled D02-1480)
and the standard deviation of the ILS data.
The D02-1480 study results in TABLE 2 are the final
sulfur test data generated from the study using “D2622 (primarily), D4294,
D5453 and CGSB” sulfur test methods. ASTM D4294 is EDXRF, D5453 is UVF and CGSB
is not defined in the research report. The total number of laboratories testing
each sample is identified in TABLE 2, but a breakdown of the number of laboratories using
each test method was not provided in the research report.
It is interesting to note that while the sulfur
results are reported in this research report as indicated in TABLE 2, there are no
comments, conclusions or discussion of these sulfur results in this research
report.
DISCUSSION
It is unclear how the B100 D6751 specification
arrived at the Note X1.1 conclusion, which indicates: “…Test Method D2622
yielded falsely high results due to the presence of the oxygen in the
biodiesel. Sulfur results using Test Method D2622 were more accurate with B20
than with B100 due to the lower oxygen content of B20 …” Several inconsistencies with this conclusion
will be discussed in the following section.
Falsely high
due to the presence of oxygen. The D02-1480 research report does not say the
B100 results were falsely high due to oxygen content—this conclusion originates elsewhere. There
is no concluding statement in the research report where the study values are
compared to the inspection values, much less any statement that the B100
samples are falsely high due to oxygen content. Additionally, no investigation
was detailed in this research report of why the B20 samples were in closer
agreement with their inspection values than the B100 test samples were with their
inspection values.
Inspection
values.
For
the biodiesel and biodiesel blend samples, there is no detail in the research
report of who conducted this testing, where these values originate (single
laboratory results, round robin testing, etc.) or what test methods were used
to determine these values. These samples do not appear to be certified
reference materials, as there is no associated uncertainty or traceability
provided with the inspection values (at least as far as the research report is
concerned).
The report does indicate that the 902 conventional
diesel sample is a round robin sample provided to NEG members for testing in
January 1999 to provide comparative data for the biodiesel blends, so it is
possible that this sample may be a reference material.
However, it is unclear where the biodiesel and
B20 biodiesel blend inspection values originate. Added to the fact that it is unusual
to provide inspection test values to study participants prior to testing, it
seems unwise to use the inspection values as the true value(s) for the samples.
Then, using these values to further make an accuracy statement about the research
report data seems speculative at best.
D2622
(primarily), D4294, D5453 and CGSB sulfur test methods. The D02-1480 study
results were generated using D2622 (primarily), D4294, D5453 and CGSB sulfur
test methods. The total number of laboratories testing each sample is
identified in TABLE 2,
but a breakdown of the number of laboratories using each test method was not
provided in the research report. This is disturbing—what does “primarily” mean? This is not
reported in the research report. When dealing with statistics from multiple
elemental analysis methods, best practices today evaluate each method
separately and combine the data if statistically feasible. Due to the low
number of participants (10–16
labs), there may not be enough degrees of freedom to separate by method. What then
determines whether a specific method is fit (or not fit) for use?
Research
report data.
Closer
investigation of the D02-1480 ILS sulfur data in TABLE 2 indicates that
the participating laboratories had trouble analyzing the ultra-low sulfur biodiesel
samples (903B and 905B highlighted in TABLE 2), because the standard deviation of
the sulfur results are greater than the sulfur values themselves. This is
likely not an issue with only one or two participants because the data was
subject to the generalized extreme Studentized deviate many-outlier (GESD) procedure,
which would remove gross outlying values/participants. This problem was not
observed with biodiesel sample 904B, which had an average measurement of 31 ppm
sulfur with a standard deviation of 16 ppm, which agrees with its inspection
value of 28 ppm.
Due to bad statistics, samples 903B and 905B should
be discarded from the analysis statement made in the D6751 specification. If
these two data points are thrown out, it can be observed that the remaining
biodiesel sample is
in close agreement with the inspection value, as it is well within D2622 reproducibility
(6.2 ppm). However, a statement based on one data point is weak, so a good
compromise would be removing the statement from the D6751 specification
altogether.
ASTM
B100 PTP program.
Perhaps the best way to put this issue to rest is to obtain some impartial data
comparing the D5453 and D2622 test methods. ASTM runs a B100 Proficiency
Testing Program (PTP)6 wherein B100 samples are
sent out three times per year to laboratory participants for analysis of
approximately 24 different properties, including sulfur. Unfortunately, D2622
is not included in this PTP program, likely due to the statement regarding
D2622 in the biodiesel specification.
Prior to the publication of this paper, a
request was made to include D2622 in the B100 PTP program. Unfortunately, it
was denied as ASTM stated that the method was inappropriate for biodiesel due
to the samples being low in sulfur and high in oxygen content.
However, it is interesting to note that D7039
(another WDXRF method, specifically for MWDXRF) is included in this PTP; in a
separate paper7 written by the author, the low bias of D7039 PTP data (in orange, FIG. 1) to D5453 PTP
data (in blue, FIG. 1)
is discussed, and it is proposed that this bias is due to the lack of oxygen
correction by participants.
Note: This low bias is the
same expectation that exists for high oxygen-containing samples in Section 6.2
of D2622 but contradicts the expectation of Note X1.1 in D6751.
Takeaways. A contradiction exists
between the ASTM D6751 biodiesel specification and the D2622 sulfur by WDXRF test
method. The B100 specification states that the high oxygen of B100 samples
leads to high bias by D2622, but the D2622 method states that high-oxygen
samples lead to sulfur Kα radiation absorption and a low sulfur bias (if not
corrected for). There is an existing research report D02-1480 detailing the D6751
specification references; however, when investigated, a number of issues include:
Although D2622 is not in the ASTM B100
proficiency testing program, comparison of another WDXRF method (D7039 sulfur
by MWDXRF) with D5453 (sulfur by UVF) indicates that the D2622 method is
correct, and that high-oxygen content causes low rather than high bias.
Considering these discoveries, an investigation
by ASTM should be undertaken to address this discrepancy. At a minimum, it is
recommended that the high bias statement should be immediately balloted for
removal from the D6751 B100 specification and D2622 be added to the ASTM B100
PTP program to collect impartial method comparison data. HP
LITERATURE CITED
LESLIE MCHENRY is an Applications Supervisor for XOS. With more than two decades of experience in petroleum applications—first as a chemist specializing in downstream product analysis at a North American petroleum refinery, then as an Applications Scientist and Supervisor at XOS—McHenry is recognized as an industry expert. She is the author of numerous white papers and ASTM XRF method revisions, including D7039, D7536, D7757 and the addition of Procedure C for XRF to ASTM D4929 “Determination of chloride content in crude oil.” As a former XOS customer, McHenry enjoys assisting others with their elemental analysis testing challenges.