This document describes a single-laboratory validation study of a combustion method for determining total sulfur content in fertilizers. The method involves combusting samples at 1150°C, converting sulfur species to SO2, and measuring SO2 levels using a thermal conductivity detector. Eleven fertilizer samples with sulfur content ranging from 1.47-91% were analyzed over multiple days. Recoveries ranged from 94.3-125.9% and precision was good, suggesting the method is accurate and precise for determining total sulfur in fertilizers. Further collaborative study is recommended to validate the method.

1. Bernius et al.: Journal of AOAC International Vol. 97, No. 3, 2014 731
Determination of Total Sulfur in Fertilizers by High
Temperature Combustion: Single-Laboratory Validation
Jean Bernius
elementar Americas, Mt. Laurel, NJ 08054
Sabine Kraus
Elementar Analysensysteme GmbH, Hanau, Germany
Sandra Hughes
elementar Americas, Mt. Laurel, NJ 08054
Dominik Margraf
Elementar Analysensysteme GmbH, Hanau, Germany
James Bartos and Natalie Newlon
Office of Indiana State Chemist, Purdue University, 175 S. University St, West Lafayette, IN 47907-2063
Hans-Peter Sieper
Elementar Analysensysteme GmbH, Hanau, Germany
Guest edited as a special report from the AOAC Agricultural
Community on “Collaborations in New and Improved Methods of
Analysis for Plant Food Materials” by Nancy Thiex.
Corresponding author’s e-mail: j_bernius@elementar-inc.com
DOI: 10.5740/jaoacint.13-385
SPECIAL GUEST EDITOR SECTION
A single-laboratory validation study was conducted
for the determination of total sulfur (S) in a variety
of common, inorganic fertilizers by combustion. The
procedure involves conversion of S species into
SO2 through combustion at 1150°C, absorption then
desorption from a purge and trap column, followed
by measurement by a thermal conductivity detector.
Eleven different validation materials were selected
for study, which included four commercial fertilizer
products, five fertilizers from the Magruder Check
Sample Program, one reagent grade product, and one
certified organic reference material. S content ranged
between 1.47 and 91% as sulfate, thiosulfate, and
elemental and organically bound S. Determinations
of check samples were performed on 3 different days
with four replicates/day. Determinations for non-
Magruder samples were performed on 2 different
days. Recoveries ranged from 94.3 to 125.9%. ABS SL
absolute SD among runs ranged from 0.038 to 0.487%.
Based on the accuracy and precision demonstrated
here, it is recommended that this method be
collaboratively studied for the determination of
total S in fertilizers.
S
ulfur (S) deficiency in corn and other crops is becoming
more prevalent. Less S is being deposited into the soil due
to reductions in power plant emissions and increased use
of low S diesel fuel. Changing agronomic practices such as no or
low tilling and early planting also contribute to S deficiency (1).
Increased use of S-containing fertilizers has driven demand
for more S determinations, including sulfate and elemental S in
blended fertilizers (Hall, B., Mosaic Fertilizer, 2013, personal
communication).
The current AOAC Method 980.02 (2) for S in fertilizer is
a gravimetric technique using barium to precipitate the sulfate
and carbon disulfide to solubilize the elemental S. This method
is labor-intensive and requires handling of hazardous materials.
As an alternative technique, combustion elemental analysis
is automated and is a safe technique for S determination. The
required instrumentation is already in use for nitrogen (N)
determination in fertilizer, AOAC Method 993.13 (3), and is
familiar to laboratories making these determinations.
Combustion analyzers without absorption/desorption
chromatography were investigated by the Office of the Indiana
State Chemist (OISC) as an alternative to the gravimetric method.
Matrix and concentration sensitivities posed several analytical
challenges, especially when S determination for a wide range
of fertilizer products or an unknown source was required. Burn
rates, scrubber types, and the need for accelerants was a function
of the fertilizer type. As a result, different scan shapes and areas
were generated for different products, and the most accurate data
were obtained when sample and calibration sources matched.
For example, a pure ammonium sulfate [(NH4)2SO4] fertilizer
source has a quick burn that requires no accelerant, whereas a
blended fertilizer product containing potassium-magnesium
sulfate (K2MgO8S2) and potassium chloride (KCl) is a slow burn
that may require an accelerant and a halogen scrubber. Another
requirement was the separation of samples into low and high S
concentrations because low concentration required a longer cell
length that could be overwhelmed by high concentration samples
resulting in carryover for the subsequent samples. As a result,
samples were presorted into categories based upon S source, S
concentration, and matrix to best match the calibration material
and concentration range. If the S source was unknown, the sample
was pretested and the scan compared to scans from various
fertilizer sources to determine the best calibration material and
concentration. For a company or laboratory that knows the source
and history of the product, presorting of samples presents no
problem but can be a challenge for a laboratory testing largely
unknown samples.
The objective of this work was to investigate an automated,
matrix-independent method using less hazardous reagents than
the current regulatory procedure that allows equally simple S
Copyright: © 2014 AOAC INTERNATIONAL.
This is an open-access article, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.

2. 732 Bernius et al.: Journal of AOAC International Vol. 97, No. 3, 2014
determination in single species and complex blended fertilizer
products.
Experimental
Principle
The sample is wrapped in tin foil and dropped into the analyzer
through a blank-free helium-purged ball valve. Oxygen is jet
injected over the sample at 1150°C. Separation of combustion
gases is performed using a thermal desorption purge and trap
chromatographic method. Combustion gas components CO2,
H2O, and SO2 are adsorbed on three specific columns. N is not
adsorbed and flows directly to the thermal conductivity detector
(TCD). Based on the detector reading, gas components are
released sequentially from their individual adsorption/desorption
columns. Time of analysis for C, N, and S is 8 min/sample. S
analysis is not available as a single analyte.
Apparatus
(a) Analytical balance.—Resolution to 0.01 mg (Mettler-
Toledo, Columbus, OH).
(b) vario MACRO cube analyzer.—Elementar
Analysensysteme GmbH (Hanau, Germany).
(c) Tin foil and tin capsule sample containers.—Elementar
Analysensysteme GmbH.
(d) Sample hand press for powdered samples.—Elementar
Analysensysteme GmbH.
(e) Sample sealing press for liquid samples.—Elementar
Analysensysteme GmbH.
Reagents
(a) Helium.—99.995% purity (Praxair, Danbury, CT).
(b) Oxygen.—99.995% purity (Praxair).
(c) Tungsten (VI) oxide (WO3) granulate.—Supplied by the
instrument manufacturer (Elementar Analysensysteme GmbH).
(d) Copper wires.—Supplied by the instrument manufacturer
(Elementar Analysensysteme GmbH).
(e) Corundum balls.—Supplied by the instrument
manufacturer (Elementar Analysensysteme GmbH).
(f) Quartz wool.—Supplied by the instrument manufacturer
(Elementar Analysensysteme GmbH).
(g) Silver wool.—Supplied by the instrument manufacturer
(Elementar Analysensysteme GmbH).
(h) Sicapent®
with indicator.—Supplied by the instrument
manufacturer (Elementar Analysensysteme GmbH).
(i) Sulfanilamide.—Sigma-Aldrich (St. Louis, MO) as sample
additive.
Test Sample Preparation
(a) Source of materials.—For method validation experiments,
S-containing materials were obtained from commercial sources,
the Magruder Check Sample Program, and previously analyzed
fertilizer samples provided by the OISC and Kentucky Division
of Regulatory Services (Lexington, KY). The materials were
chosen to represent a range of S concentrations and matrixes that
make up 90% of fertilizer tonnage worldwide (Hall, B., Mosaic
Fertilizer, personal communication, 2013). The concentration
of the validation materials ranged from 1.47 to 91% S and
included sulfate, thiosulfate, and blended elemental S and sulfate-
containing fertilizers. To obtain materials for Magruder Check
Sample Program go to www.magruderchecksample.org.
(b) Sampling and sample preparation.—The liquid
ammonium thiosulfate sample was collected using AOAC
Method 969.01 (4). A field inspector collected approximately
500 mL sample from a tap point on a liquid bulk storage tank.
After shaking the plastic field container well, approximately
50 mL was poured into a clean 100 mL glass jar that served as the
laboratory test portion. The remaining sample was retained for
result confirmation or dispute resolution. The solid elemental S
sample was collected using AOAC Method 929.01 (5). Multiple
cores were taken from a bulk sample resulting in approximately
a 2.3 kg subsample. The subsample was reduced further with a
Carpco (Jacksonville, FL) SS16-25 gated riffler. Approximately
200 g sample was finely ground using a Retsch (Newtown, PA)
ZM200 grinder and a 1 mm screen and then poured into a clean
100 mL glass jar. The glass jar was rotated multiple times to
ensure sample uniformity before testing. An unground portion
was retained for result confirmation or dispute resolution. All
samples were stored at room temperature in restricted access,
locked storage.
The Magruder samples were used as prepared by the Check
Sample Program supplier. All of these samples were stored at
room temperature in sealed plastic pouches.
Calibration Standard
Factory calibration of the combustion analyzer was tested daily
using 99% purity sulfanilamide. Drift correction was applied by
the instrument software.
Operating Procedure
The elements C, H, N, and S bound in the sample burn to
form CO2, H2O, N2, NOX, SO2, and SO3. Halogen bound in the
sample reacts to form volatile halogen compounds within the
combustion tube. WO3 granulate delivers additional oxygen to
prevent the formation of nonvolatile sulfates and to bind alkali
and alkaline earth elements. The carrier gas flow transfers the
gaseous combustion products into the reduction tube, where
copper contact completely reduces NOX to N2. SO3 is reduced to
SO2. Volatile halogen compounds are bound on silver wool. H2O
contained in the reaction gas mixture is bound in an absorption
tube filled with Sicapent.
N2 gas bypasses the adsorption/desorption columns and passes
directly to the TCD. When the N2 peak comes down to baseline,
Table 1. Operating conditions
Combustion furnance temp., °C 1150
Reduction furnace temp., °C 850
Oxygen pressure, bar 2.0–2.2
Helium pressure, mbar 1200–1250
Method for oxygen dosing Sulf1, Sulf2, blank according to
instrument software

3. Bernius et al.: Journal of AOAC International Vol. 97, No. 3, 2014 733
the adsorption column is heated to CO2 and then SO2 desorption
temperatures. The CO2 and H2O adsorption columns are bypassed
via two valves. An absorption tube removes traces of H2O from
the gas stream. SO2 is desorbed and enters the detector with the
carrier gas. A fan cools the adsorption column, making it ready
for the next analysis.
The software reports C, N, and S but since C is not of interest,
analysts can disregard C results. N results may be of interest. The
N method conforms to AOAC Method 993.13 (3).
Caution: During operation, the furnaces, the valve area, and
the quartz glass bridge inside the instrument reach very high
temperatures (in excess of 1000°C). Even after switching off the
instrument, these components stay hot for long periods of time.
Serious burns can occur if working inappropriately inside the
instrument.
(a) Weigh 26–50 mg samples into tin foil cones or tin capsules.
Add between 260 and 500 mg WO3 combustion aid for a 10:1
ratio of aid to sample. Seal with appropriate sealing device.
(b) Turn on the vario MACRO cube analyzer. Conduct a leak
check. Ensure combustion, reduction, and drying tubes do not
need replacement. When furnace temperatures are up to required
values, analysis may begin.
(c) Load carousel with start-up samples: three blanks,
three run in or conditioning samples using sulfanilamide, and
three standard samples using 15 mg sulfanilamide. Include a
sulfanilamide standard sample after every 10 samples and as the
last sample in the batch to check for drift in the calibration.
(d) Analyze standards and samples while monitoring
instrument conditions for proper gas flows, restrictions, leaks,
and consumable reagent life.
Operating Conditions
See Table 1.
Calculations
Calibration is performed by analyzing samples of
nonhygroscopic fine chemicals with known concentrations
(standard samples) in which the measured peak areas are
correlated with the corresponding absolute element content. In
a standard sample, the following parameters are known: element
concentration (content, %) and weight. Absolute element content
is computed according to the following equation:
A = WC/100
where A = absolute element content in mg, W = sample weight in
mg, and C = element concentration in percentage.
Quality Control
(a) Blank determination should be performed before daily
operation and any time a reagent or carrier gas tank is replaced.
(b) Drift correction using 15 mg sulfanilamide standard should
be performed before each day’s analyses. If the drift correction or
daily factor exceeds 0.9 or 1.1, maintenance must be performed
to check for leaks or that reagents in the combustion or reduction
tubes are not spent.
(c) Include one reference material with each batch of 30
samples. Results should be within limits specified for reference
material. Reference materials can include a fine chemical of
known S content such as sulfanilamide, a Magruder check
sample, or a National Institute of Standards and Technology
reference material.
(d) Perform replicate analysis on at least 10% of the samples,
or every sample if homogeneity is difficult to achieve. Replicate
results should be within 10–15% of the mean value of the
replicates.
Validation of the Method
This method was validated in a single laboratory by a
single analyst. Validation experiments were conducted using
11 materials including five from the Magruder Fertilizer Check
Sample Program, four previously analyzed fertilizer samples
from the OISC and Kentucky Division of Regulatory Services,
and two commercial S-containing materials. For the Magruder
samples, the grand average of results, after outlier removal,
obtained for each analyte/method combination on a check sample
material was considered a consensus value. The consensus
values were used in the absence of reference materials to validate
the method. Participating laboratories submitted their results
to the Association of American Plant Food Control Officials
Magruder Program, and the grand average (minus outliers) for
each analyte/method combination was publicly posted. Label
values for ammonium sulfate U.S. Pharmacopeia grade and
O-acetyl-l-serine cysteine were used. Previously determined S
concentrations were used for ammonium thiosulfate 12-1591,
sulpomag 12-1773, and elemental S 12-0508 as determined by
the OISC. Previously determined S concentration was used for
potassium sulfate 12289 as determined by the Kentucky Division
of Regulatory Services (Table 2).
(a) Linearity and range.—The instrument was calibrated
according to the manufacturer’s instruction utilizing a total of
32 different sulfanilamide (purity ≥99%) and phenylalanine
(purity ≥99%) samples of known weight. The calibration covered
19.73–11 190.62 µg absolute element content for S. For a 30 mg
sample, this correlates to 660 ppm to 37.3% S. The calibration
functions for S content are best described on the basis of fourth-
order polynomial functions. SD of the method for S was 11.73 µg.
(b) Accuracy.—Matrix effects were investigated using three
sulfanilic acid samples with an average sample weight of 30 mg.
A recovery of >99% was achieved.
The accuracy of the method was determined by comparing
results to the accepted published values of the check samples,
previously analyzed samples, and label values. The accepted
value of the check sample (usually called the grand average) was
calculated from results submitted to the respective check sample
program managers from the dozens of laboratories participating,
using statistical packages established by the Magruder Program.
The difference between the grand average or previously reported
results and the results obtained by the proposed method was
reported as both a recovery and a bias, where a negative value
indicates the method average is lower than the grand average and
a positive number indicates the method average is higher than
the grand average. Recovery and bias were calculated for each
validation material as compared to the published available S
grand averages (Table 3).
(c) Precision.—Each Magruder check sample validation
material was analyzed on 3 different days with four replicates/
day. Other materials were analyzed on 2 different days with

4. 734 Bernius et al.: Journal of AOAC International Vol. 97, No. 3, 2014
four replicates/day. Within-run repeatability and among-run
repeatability of the method was expressed as the SD. Within-run
repeatability includes variability associated with the instrument.
Among-run repeatability includes all variables associated with
the method (Table 4).
(d) LOD and LOQ.—As suggested in theAOAC guideline (6)
for single-laboratory validation (SLV), the LOD was calculated
as the blank value plus three times the SD of the blank. The
LOQ was calculated as the blank value plus 10 times the SD of
the blank. LOQ is also referred to as “limit of determination.”
LOD and LOQ of the proposed method were 47 and 106 µg
S, respectively. For comparison purposes, the lowest absolute
amount of S in the validation materials was in Sample 1. When
we analyzed 50 mg of Sample 1 (with 1.56% S), the absolute S
value was 780 µg, which exceeded the LOQ of the method.
(e) Ruggedness trial.—A ruggedness trial with only eight
determinations was used to explore the effects of the seven
most important factors as described in the AOAC requirements
for SLVs (6): no blank measurement, improper WO3 addition,
incorrect oxygen dosing, not wrapping sample in tin, improper
combustion temperature, high and low sample weights, and
broken oxygen dosing device. Each factor was assigned with
reasonable high and low values. The effect of each factor was
determined by the difference in S content when a factor was
changed from high to low levels. Factors with extreme difference
between high and low levels were noted.
Magruder 2012-10 was the only matrix used in the trial. This
material was selected as the most recent material available at the
time of the trial. Source of S in Magruder 2012-10 is ammonium
sulfate with an expected S value of 5.48%.
The squared differences of each parameter were determined
according to the Youden and Steiner Statistical Manual of the
AOAC (7) and listed based on their value (Table 5).
Discussion
Accuracy
Table 3 shows the comparison of the elementar vario Macro
CUBE method to the known values for S. Recovery for all
materials averaged between 94.29 and 125.94% S. Average
recovery for S was 101.07%. Bias for S ranged between –1.082
and 2.013%. Recoveries for Samples 1–3 appear to be low,
but they were within the error range for the current accepted
method. At 125.94%, recovery of Sample 5 Magruder 2011-07
ran higher than expected but was within the range of laboratories
that reported results to the program. All Magruder check sample
S results of the proposed method agreed with those obtained
Table 2. Source of validation materials
No. Sample Expected S, % Source of material Source/comments
1 2011-05 1.56 Magruder Processed poultry manure complex forms of S
2 2011-08 6.39 Magruder Turf fertilizer blend, both elemental and sulfate S
3 2011-06 11.95 Magruder Lawn and garden blend, all sulfate S
4 2011-01 13.22 Magruder Lawn and garden blend, all sulfate S
5 2011-07 7.76 Magruder Homogenous agriculture fertilizer, 5% sulfate and
5% elemental S
6 (NH4)SO4 24 USP Fisher Scientific Homogenous material, used in alkaline soils, source all sulfate S
7 S 12-508 90 OISC Raw sulfur recovered from oil refining, all elemental S
8 Cysteine 26.72 Sigma-Aldrich Analytical grade,organic form S
9 Sulpomag 12-1773 22 OISC Langbenite, mined natural mineral fertilizer/all sulfate S
10 (NH4)2SO3 12-591 26 OISC Liquid fertilizer
11 K2SO4 KY12289 18 KY Division of Regulatory Services Homogenous material/all sulfate S
Table 3. Accuracy (bias and recovery) determined by comparison to validation materials S content
Sample % S Grand average % abs SD Grand average
a
% S Proposed method % abs Repeatability Bias Recovery, %
1 1.56 0.12 1.471 0.022 –0.089 94.29
2 6.39 0.38 6.202 0.159 –0.188 97.05
3 11.95 0.51 11.742 0.118 –0.208 98.25
4 13.22 0.68 13.265 0.207 0.045 100.34
5 7.76 2.50 9.773 0.082 2.013 125.94
6 24 Not known 23.964 0.087 –0.036 99.85
7 90 Not known 91 0.246 1.000 101.11
8 26.72 Not known 26.817 0.320 0.097 100.36
9 22 Not known 22.044 0.543 0.044 100.20
10 26 Not known 24.918 0.198 –1.082 95.83
11 18 Not known 17.755 0.138 –0.245 98.63
a
% abs = Absolute SD within runs.

5. Bernius et al.: Journal of AOAC International Vol. 97, No. 3, 2014 735
according to AOAC Method 980.02 (2) within the experimental
error.
Ruggedness
Ruggedness trial results showed no appreciable effect from the
deviations of the method as noted, although caution should be
taken to determine a blank measurement and use the proper 10:1
ratio of combustion aid to sample weight.
Precision
Within-run precision was calculated for all validation materials.
In Table 4, SD values for within-run precision ranged from 0.013
to 0.424% abs. Samples that exhibited the widest spread between
Day 1 and Day 2 were Sample 7 elemental S, Sample 8 cysteine,
and Sample 9 sulpomag. Sample 4 lawn and garden blend with all
sulfate varied the most of the Magruder samples.
In Table 4, abs SL among-run precision ranged from 0.038
to 0.487. Sample 1 Magruder 2011-05 exhibited the tightest
values. Sample 9 sulpomag, a mined natural fertilizer containing
Langbenite, had the widest values possibly due to a high melting
point.
Conclusions
Arapid method for the determination of S and S/sulfate blended
fertilizers using high temperature combustion with adsorption/
desorptioncolumnswasvalidatedasfitforthedesignatedpurpose.
The method was shown to be accurate at 1% S concentration.
The precision of the method was demonstrated to be acceptable.
The method proved to be time- and resource-efficient, allowing
analysis of as many as 50 samples/8 h day with a single
calibration curve and much fewer hazardous reagents than the
current procedure.
Acknowledgments
We thank Bill Hall (Mosaic Fertilizer, Lakeland, FL) and
Melton Bryant (retired, University of Kentucky, Lexington, KY)
for guidance; April Holcomb (Agilent Technologies, Baltimore,
MD) for analytical support; and Mark Larson (elementar
Americas, Mt. Laurel, NJ) for editing support.
References
(1) Camberato, J. (2012) Soil Fertility Update, Purdue University
Department of Agronomy. http://www.king.corn.org/news/
timeless/SulfurDeficiency.pdf (accessed November 11, 2013)
(2) Official Methods of Analysis (2005) 18th Ed., AOAC
INTERNATIONAL, Gaithersburg, MD, Method 980.02
(3) Official Methods of Analysis (2005) 18th Ed., AOAC
INTERNATIONAL, Gaithersburg, MD, Method 993.13
(4) Official Methods of Analysis (2005) 18th Ed., AOAC
INTERNATIONAL, Gaithersburg, MD, Method 969.01
(5) Official Methods of Analysis (2005) 18th Ed., AOAC
INTERNATIONAL, Gaithersburg, MD, Method 929.01
(6) AOAC Guidelines for Single-Laboratory Validation of Chemical
Methods (2002) AOAC INTERNATIONAL, Gaithersburg, MD
(7) Youden, W.J. Steiner, E.H., (1975) Statistical Manual of the
AOAC, AOAC INTERNATIONAL, Gaithersburg, MD
Table 4. Precision of the method
Abs SD
Sample Expected recovery S, % Average recovery, % Abs SL
a
Day 1
b
Day 2 Day 3 HorRat
1
c
1.56 94.29 0.038 0.028 0.018 0.018 0.39
2 6.39 97.05 0.147 0.190 0.137 0.146 0.84
3 11.95 98.25 0.13 0.069 0.146 0.126 0.36
4 13.22 100.34 0.219 0.189 0.297 0.069 0.57
5 7.76 125.94 0.171 0.058 0.077 0.104 0.29
6
d
24 99.85 0.174 0.029 0.051 0.139 0.14
7 90 101.11 0.235 0.013 0.347 — 0.13
8 26.72 100.36 0.323 0.165 0.424 — 0.48
9 22 100.2 0.487 0.763 0.088 — 0.98
10 26 95.83 0.185 0.268 0.082 — 0.32
11 18 98.63 0.128 0.144 0.132 — 0.29
a
Abs SL defined as absolute SD among runs or within-laboratory precision.
b
Abs SD defined as absolute SD within runs.
c
Twelve replicates for Samples 1–5: four replicates run on 3 days.
d
Eight replicates for Samples 6–11: four replicates run on 2 days.
Table 5. Ruggedness trial method deviations and squared
differences
Parameter Squared differences
No blank measurement 0.08
WO3 addition 8:1, 12:1 0.06
Wrong method oxygen dose 0.03
Wrong sample wrap, not tin 0.003
Combustion 1100 or 1170°C 0.002
Sample weight 20/60 mg 0.002
Broken oxygen dosing device 0.00001