502 ANDERSON: JOURNAL OF AOAC INTERNATIONAL VOL. 92, NO.
2, 2009
FOOD COMPOSITION AND ADDITIVES
Determination of Bromine in Regulated Foods with a
Field-Portable X-Ray Fluorescence Analyzer
DAVID L. ANDERSON
U.S. Food and Drug Administration, Center for Food Safety and Applied Nutrition, Chemical Contaminants Branch,
HFS-716, 5100 Paint Branch Pkwy, College Park, MD 20740-3835
A field-portable X-ray fluorescence analyzer,
factory-calibrated for soil analysis, was used to
measure bromine (Br) mass fractions in reference
materials, flour, bakery products, malted barley,
selected U.S. Food and Drug Administration Total
Diet Study foods, and other food products. By
using a calibration based on instrumental neutron
activation analysis results for Br in reference
materials, accurate quantitative results, confirmed
by z-scores, could be obtained for mass fractions
of about 2–55 mg/kg. These results confirmed
accuracy of results (with larger uncertainties)
obtained by applying a simple correction factor to
the analyzer’s output value. Results showed that
very short analysis times (<2 min) would be needed
to screen foods for Br content at regulatory levels
for brominated and enriched brominated flour
(24 mg/kg Br) and whole wheat flour and bakery
products (36 mg/kg Br). Feasibility for
determination of Br in malted barley at the
regulatory level (75 mg/kg Br) was demonstrated,
but quantitative results at that level could not be
assured because no reference material with a
suitable mass fraction was available. Br mass
fractions for all foods tested were well below
regulatory levels.
P
otassium bromate (KBrO3), considered to be a
carcinogen for animals (1, 2), is used as an additive to
flour as a maturing agent and for dough
conditioning (3). KBrO3 is also used in the malting process for
barley to reduce rootlet growth, thus reducing heat generation
(eliminating refrigeration costs), germination area, and
soluble nitrogen content for the mash (4). Although bromate is
reduced to the safe bromide form in properly baked products,
residual bromates have been found in bread products due to
improper baking conditions or excessive bromation. Previous
studies (3) have shown that total bromine (Br) concentrations
in bakery products vary linearly with the amount of bromate
Received July 18, 2008. Accepted by SG September 18, 2008.
Corresponding author’s e-mail: david.anderson@fda.hhs.gov
added to the flour component regardless of the species present
in the final product.
The U.S. Food and Drug Administration (FDA) regulates
the amount of KBrO3 that may be added to cereal flours (5),
bakery products (6), and the barley malting process (7). The
amounts that may not be exceeded are 50 mg/kg
(corresponding to 24 mg/kg Br) for brominated and enriched
brominated flour and 75 mg/kg (corresponding to 36 mg/kg
Br) for brominated whole wheat flour and bakery products
(regular, enriched, milk, raisin, and whole wheat bread, rolls,
and buns). The amount added to the barley malting process
may not exceed 75 mg/kg, in this case, measured as Br. Malted
barley with added KBrO3 may only be used for fermented and
distilled beverages. Residue of inorganic bromides in
fermented malt beverages may not exceed 25 mg/kg (as Br).
In the present work, the potential of X-ray fluorescence
(XRF) spectrometry to measure total Br mass fractions in
grain-based food products was studied for the purpose of
developing rapid screening and quantitative methods.
Previous FDA studies (8–10) with radioisotope- and X-ray
tube-based XRF analyzers showed that many matrixes,
including ceramics, soldered cans, food wraps, beverages, and
solid foods could be successfully screened for potentially
toxic elements such as As, Cd, Hg, and Pb. In this work, a
field-portable X-ray tube-type analyzer was used to determine
Br over mass fraction ranges relevant to regulatory level mass
fractions in reference materials (RMs), including National
Institute of Standards and Technology (NIST) Standard
Reference Materials (SRMs), and in selected FDA Total Diet
Study (TDS; 11) foods and flour, grain, malted barley, and
other food products. Although the portable analyzer was not
developed (or factory-calibrated) for the purpose of analyzing
food matrixes, results can be obtained by using the
instrument’s soil matrix analysis mode. In the present study,
simple methods of data correction were devised, and corrected
results compared to known values and results obtained by
instrumental neutron activation analysis (INAA; 12, 13) for
SRMs and food products.
Experimental
Principle
A miniature X-ray tube inside the instrument accelerates
electrons from a cathode to strike a silver metal anode. X-rays
in the energy range of about 10–35 keV (maximum intensity
Table 1. X-ray fluorescence analyzer output and calculated results for Br in reference materials compared to NIST Certificate, INAA, and consensus valuesa
Analyzer output, ppm INAAd Consensuse

Calculated from factor, Calculated from NIST Certificate,

Reference material Raw Blank-correctedb mg/kgc fit, mg/kg mg/kg mg/kg z mg/kg z

One analysis
f
SRM 1566 Oyster Tissue 78.19 ± 1.50 75.62 ± 1.50 52 ± 9 54.6 ± 2.2 55 54.6 ± 4.3 0.0 54 ± 5 0.1
f
SRM 1567 Wheat Flour 13.74 ± 0.70 11.17 ± 0.70 9±2 9.0 ± 1.1 9 9.0 ± 1.2 0.0 8.7 ± 0.8 0.3
SRM 1568 Rice Flourf 3.30 ± 0.52 0.73 ± 0.52 2±1 0.6 ± 0.9 1 0.63 ± 0.19 0.0 0.64 ± 0.12 0.0
f
SRM 1570a Spinach 49.32 ± 1.33 46.75 ± 1.33 33 ± 6 35.5 ± 2.0 — 35.5 ± 2.3 0.0 43 —
SRM 1567a Wheat Flour 11.10 ± 0.64 8.53 ± 0.64 7±2 6.9 ± 1.0 — 6.4 ± 0.5 0.9 — —
SRM 1568a Rice Flour 11.97 ± 0.66 9.40 ± 0.66 8±2 7.6 ± 1.1 — 7.9 ± 0.5 –0.5 — —
SRM 1570 Spinach 77.90 ± 1.73 75.33 ± 1.73 52 ± 9 54.4 ± 2.5 — 50.7 ± 7.8 0.9 — —
SRM 1577a Bovine Liver 15.06 ± 0.79 12.49 ± 0.79 10 ± 2 10.1 ± 1.3 9 — — 9.7 ± 1.2 0.3
SRM 1577b Bovine Liver 16.55 ± 0.80 13.98 ± 0.80 11 ± 2 11.2 ± 1.3 9.7 — — 9.6 —
SRM 1573 Tomato Leaves 28.79 ± 1.09 26.22 ± 1.09 19 ± 4 20.6 ± 1.7 26 — — 22 ± 2 –0.6
SRM 1515 Apple Leaves 3.44 ± 0.58 0.87 ± 0.58 2±1 0.7 ± 1.0 1.8 1.85 ± 0.05 –2.3 1.76 ± 0.13 –2.0
SRM 1547 Peach Leaves 17.80 ± 0.87 15.23 ± 0.87 12 ± 2 12.2 ± 1.4 11 12.34 ± 0.67 –0.2 10.9 ± 0.6 1.4
SRM 1549 Milk Powder 16.63 ± 0.82 14.06 ± 0.82 11 ± 2 11.3 ± 1.3 12 11.9 ± 0.4 –0.9 11.4 ± 1 –0.1
SRM 2710 Montana Soil <13.70 — <9 — 6 — — — —
SRM 2711 Montana Soil 8.62 ± 1.93 6.05 ± 1.93 6±3 4.9 ± 3.1 5 — — — —
SiO2 blank <2.4 — <1 — — — — — —
Deionized water <1.4 — <1 — — — — — —
Multiple analyses of same portion

SRM 1566 Oyster Tissue 9 (n = 2) 76.04 ± 5.59 73.47 ± 5.59 51 ± 12 53.3 ± 8.1 55 54.6 ± 4.3 –0.3 54 ± 5 –0.1
SRM 1571 Orchard Leaves (n = 5) 13.40 ± 0.96 10.83 ± 0.96 9±2 8.7 ± 1.6 10 9.96 ± 0.15 –1.6 9.5 ± 0.9 –0.6
SRM 1572 Citrus Leaves (n = 3) 11.41 ± 0.33 8.84 ± 0.33 7±1 7.2 ± 0.5 8.2 8.37 ± 0.61 –3.0 8.2 ± 0.4 –2.2
SRM 1577 Bovine Liver (n = 3) 16.09 ± 0.94 13.52 ± 0.94 11 ± 2 10.9 ± 1.5 — — — 9.1 ± 1.0 1.4
CFSAN Cocoa Powder (n = 4) 47.35 ± 1.02 44.78 ± 1.02 31 ± 6 34.1 ± 1.6 — 32.9 ± 1.3 1.2 — —

a
Analyzer output uncertainties are ±1s (±1 SD for multiple analyses). Calculated and instrumental neutron activation analysis (INAA) uncertainties are ± 2s. National Institute of Standards and
Technology (NIST) Certificate values are informational.
b
Blank = 2.57 ± 0.04 ppm (Figure 2).
c
Factor = 0.67 ± 0.11 (Figure 1).
d
Ref. 11 and unpublished data; z compared to fit calculation.
e
Uncertainties are ± 1 SD as given in ref. 16; z compared to fit calculation.
f
Used to determine blank and sensitivity.
ANDERSON: JOURNAL OF AOAC INTERNATIONAL VOL. 92, NO. 2, 2009 503
504 ANDERSON: JOURNAL OF AOAC INTERNATIONAL VOL. 92, NO. 2, 2009

Table 2. X-ray fluorescence results for Br in FDA Total Diet Study foods compared to INAA measurementsa

Analyzer output, ppm INAA

Calculated from Calculated

Total Diet Study foodb Raw Blank-corrected factor, mg/kg from fit, mg/kg mg/kg z

2003-1 317 Teething biscuits 4.45 ± 0.55 1.88 ± 0.55 3±1 1.6 ± 0.9 2.0 ± 0.1 –0.9
2004-3 178 Chocolate cake 3.44 ± 0.57 0.87 ± 0.57 2±1 0.7 ± 1.0 2.5 ± 0.2 –3.6
2004-3 183 Chocolate chip cookie 8.92 ± 0.60 6.35 ± 0.60 6±1 5.2 ± 1.0 6.7 ± 0.4 –2.9
2004-3 184 Sandwich cookie 4.92 ± 0.56 2.35 ± 0.56 3±1 1.9 ± 0.9 1.5 ± 0.1 0.9
2004-3 185 Apple pie 3.57 ± 0.63 1.00 ± 0.63 3±1 0.8 ± 1.1 2.0 ± 0.1 –2.2
2004-3 186 Pumpkin pie 3.26 ± 0.52 0.69 ± 0.52 2±1 0.6 ± 0.9 1.8 ± 0.1 –2.7
2005-2 290 Donut 4.17 ± 0.54 1.60 ± 0.54 3±1 1.3 ± 0.9 2.0 ± 0.6 –1.2
2005-2 291 Brownie 7.27 ± 0.62 4.70 ± 0.62 5±1 3.8 ± 1.0 3.7 ± 0.5 0.3
2005-2 292 Sugar cookie 2.52 ± 0.49 –0.05 ± 0.49 2±1 0.0 1.2 ± 1.0 —
2005-2 294 Pretzel 3.94 ± 0.58 1.37 ± 0.58 3±1 1.1 ± 1.0 3.1 ± 2.7 –1.3
2007-3 063 Flour tortilla 8.95 ± 0.64 6.38 ± 0.64 6±1 5.2 ± 1.0 NA —
2007-3 187 Chocolate candy 9.94 ± 0.66 7.37 ± 0.66 7±1 6.0 ± 1.1 NA —
2007-3 331 Meal replacement 4.85 ± 0.54 2.28 ± 0.54 3±1 1.9 ± 0.9 NA —
2007-3 341 Refried beans 2.22 ± 0.52 –0.35 ± 0.52 1±1 –0.3 NA —
2007-3 371 Granola 2.83 ± 0.54 0.26 ± 0.54 2±1 0.2 ± 1.0 NA —

a
Analyzer output uncertainties are ±1s. Calculated and instrumental neutron activation analysis (INAA) uncertainties are ± 2s; NA = not
analyzed.
b
Market basket number, food item number, food name.

from 20 to 22 keV) are produced for irradiation of the test
portion. This excitation energy range provides superior
relative fluorescence yields and low Compton scatter
interference for elements (including As, Se, Br, Pb, and Hg)
with K or L X-ray emissions in the range of 9–14 keV.
Fluorescent Ka and Kb Br X-rays (11.91 and 13.30 keV,
respectively) from the test portion are detected by a
miniaturized electronically cooled silicon crystal X-ray
detector (Si PiN diode), and a multichannel analyzer
processes the signals to produce the spectral data (a histogram
of number of X-rays detected versus X-ray energy). Detailed
descriptions of XRF principles and analytical applications can
be found elsewhere (14, 15).
Description of Analyzer
The Innov-X Systems (9, 16, 17) Model a-6000s analyzer
used for this study may be operated either as a handheld
instrument (trigger-controlled) or situated in a testing stand
(software-controlled). For both configurations, the analyzer is
operated in conjunction with a Hewlett Packard iPAQ Pocket
PC. In testing stand mode, the analyzer may also be controlled
by connection to and via software on a Windows XP-based
PC. The analyzer is equipped with 5 factory-calibrated
fundamental parameter modes of analysis (soil, filter/thin
film, alloy, plastic, and industrial paint). For the present study,
the standard soil mode was used in the testing stand
configuration, and results were given as parts per
million (ppm) calculated from factory-set fit and calibration
routines on the iPAQ display and in the result files evaluated
after transfer to a Windows laptop PC and/or Macintosh
Power PC. A Mylar film was taped over the analyzer’s Kapton
window to prevent damage or contamination during sample
analysis. Analyses of a variety of materials in both handheld
and testing stand configurations have shown no significant
differences in results. Calibration curve analysis and data
reduction were performed on a Macintosh Power PC by using
KaleidaGraph plotting software and custom Microsoft Excel
spreadsheets.
Materials, Preparation, and Analysis
RMs and foods analyzed are listed in Tables 1–4. For
standardization and accuracy verification, 18 NIST SRMs,
Center for Food Safety and Applied Nutrition (CFSAN)
Cocoa Powder in house reference material (unpublished data),
and SiO2 powder and deionized water blanks were analyzed
(Table 1). With the exception of the deionized water, these
materials were all in the form of fine homogeneous powders.
RM portions ranging in mass from 2 to 7 g were transferred to
31 mm diameter polyethylene X-ray cells (capacity about
7 mL, corresponding to a sample depth of about 1.5 cm) and
packed tightly with a metal spatula, providing for an
effectively infinite X-ray path thickness. Mylar films were
attached to the filled cells with snap rings. Foods analyzed
(portion masses of 3–9 g) included 15 FDA TDS foods
(Table 2), 13 commercial flours, baked goods, or snack
products (Table 3), and 7 malted barley products (Table 4).
 ANDERSON: JOURNAL OF AOAC INTERNATIONAL VOL. 92, NO. 2, 2009 505

Table 3. X-ray fluorescence results for Br in flour, bread, and other grain-based productsa, analyzed for 2 min in
31 mm X-ray cell, unless noted
Analyzer output, ppm

Calculated from Calculated from fit,

Product Raw Blank-corrected factor, mg/kg mg/kg nb

Flour, brand 1 7.77 ± 0.62 5.20 ± 0.62 5±1 4.2 ± 1.0 1

Flour, brand 2 9.52 ± 0.36 6.95 ± 0.37 6.4 ± 1.2 5.7 ± 0.6 3
Flour, brand 2, in polyethylene bag 11.74 ± 0.66 9.17 ± 0.66 8±2 7.4 ± 1.1 1
White bread 11.57 ± 0.42 9.00 ± 0.42 7.6 ± 1.4 7.3 ± 0.7 3
White bread, in bag 18.82 ± 0.95 16.25 ± 0.95 13 ± 3 13.0 ± 1.5 1
White bread, crust, direct 17.13 ± 0.93 14.56 ± 0.93 11 ± 2 11.7 ± 1.5 1
White bread, interior, direct 18.98 ± 1.08 16.41 ± 1.08 13 ± 3 13.1 ± 1.7 1
Whole wheat bread 5.84 ± 0.27 3.27 ± 0.27 4.0 ± 0.8 2.7 ± 0.4 4
Whole wheat bread, crust, direct 8.36 ± 0.30 5.78 ± 0.31 5.7 ± 1.1 4.7 ± 0.5 2
Whole wheat bread, interior, direct 9.00 ± 0.76 6.43 ± 0.76 6±1 5.2 ± 1.2 1
Potato roll (hot dog) 9.01 ± 0.69 6.44 ± 0.69 6±1 5.2 ± 1.1 1
Cookie, brand 1 3.50 ± 0.49 0.93 ± 0.49 2.4 ± 0.8 0.8 ± 0.8 5
Cookie, brand 1, direct 5.72 ± 0.60 3.15 ± 0.60 4±1 2.6 ± 1.0 1
Cookie (graham) 6.58 ± 0.58 4.01 ± 0.58 5±1 3.3 ± 1.0 1
Cracker, brand 1 4.66 ± 0.55 2.09 ± 0.55 3±1 1.7 ± 0.9 1
Cracker, brand 2 6.31 ± 0.59 3.74 ± 0.59 4±1 3.1 ± 1.0 1
Cracker, brand 3 6.37 ± 0.56 3.80 ± 0.56 4±1 3.1 ± 0.9 1
Salted pretzels 4.78 ± 0.60 2.21 ± 0.60 3±1 1.8 ± 1.0 1
Tortilla chips 2.42 ± 0.52 –0.15 ± 0.52 1±1 –0.1 1
Potato chips 3.04 ± 0.53 0.47 ± 0.53 2±1 0.4 ± 0.9 1

a
Analyzer output uncertainties are ±1s. Calculated uncertainties are ±2s.
b
Repeat analyses of same portion. Factor-corrected mass fraction uncertainties include the ±2 measurement SD.

Composited TDS food, collected from 4 Market Baskets
(MBs) from fiscal years 2003–2005 and 2007, were mixed
well and transferred to X-ray cells and packed tightly with a
metal spatula to form a smooth flat surface for analysis. Foods
listed in Table 3 were similarly packed, as received, into X-ray
cells, except for the cookies, crackers, and snack foods, which
were first crushed while inside polyethylene bags. One flour
and one bread were analyzed through polyethylene bag walls
in addition to X-ray cell analysis. The breads and one brand of
cookie were analyzed directly without any packaging. Whole
grain malted barley portions were analyzed as received,
packed in X-ray cells as tightly and smoothly as possible, and
then pulverized in a coffee mill to match the RM’s matrix
more closely and reanalyzed. All test portions were analyzed
for 2 min in the testing stand mode.
INAA procedures are discussed in detail
elsewhere (12, 13, 18). Br results were obtained by
irradiating, typically, 0.3–1 g portions for either 6 h or 10 s at
thermal neutron fluence rates of 0.3 ´ 1014 or 1.1 ´
1014 cm–2/s, respectively, at the NIST Center for Neutron
Research in Gaithersburg, MD. Calibration was performed by
analyzing cellulose-based standards prepared with KBr
82
solutions for gamma-ray emissions from the isotopes Br
(half-life = 1.47 days) and 80mBr (half-life = 4.42 h).
Calibration
Calibration was performed by comparing XRF Br results
for NIST SRMS and CFSAN Cocoa Powder to known Br
mass fractions, which ranged from 0.6 to 55 mg/kg. No SRMs
are certified for Br content; only information values are
available. Therefore, Br mass fractions from published data
[INAA results for SRMs 1566, 1567, 1568, 1570a, 1567a,
1568a, and 1570 (12) and consensus values for SRMs 1577a
and 1573 (19)] and unpublished FDA INAA results were used
as true values for this study. These mass fractions agreed well
with available NIST information values (Table 1). Using these
known mass fractions, analyzer output was corrected in
2 ways.
First, a correction factor, xc/xm, was used, derived from the
average percent true value recovery (%TVR), defined as:
%TVR = 100 ´ xm/xc
506 ANDERSON: JOURNAL OF AOAC INTERNATIONAL VOL. 92, NO. 2, 2009

Table 4. X-ray fluorescence results for Br in malted barley productsa

Analyzer output, ppm

Calculated from factor, Calculated from fit,

Product Raw Blank-corrected mg/kg mg/kg

Honey Malt, pulverized 9.20 ± 0.65 6.63 ± 0.65 6±1 5.4 ± 1.1
Honey Malt, whole grains 11.63 ± 0.68 9.06 ± 0.68 8±2 7.3 ± 1.1
Crystal Malt, pulverized 6.27 ± 0.59 3.70 ± 0.59 4±1 3.0 ± 1.0
Crystal Malt, whole grains 7.35 ± 0.63 4.78 ± 0.63 5±1 3.9 ± 1.0
Plain Malt, pulverized 12.10 ± 0.68 9.53 ± 0.68 8±2 7.7 ± 1.1
Plain Malt, whole grains 9.88 ± 0.66 7.31 ± 0.66 7±1 5.9 ± 1.1
Toasted Wet Malt, pulverized 9.46 ± 0.63 6.89 ± 0.63 6±1 5.6 ± 1.0
Toasted Wet Malt, whole grains 10.38 ± 0.65 7.81 ± 0.65 7±1 6.3 ± 1.1
Toasted Dry Malt, pulverized 9.79 ± 0.65 7.22 ± 0.65 7±1 5.9 ± 1.1
Toasted Dry Malt, whole grains 15.35 ± 0.75 12.78 ± 0.75 10 ± 2 10.3 ± 1.2
Malt extract syrup 13.27 ± 0.73 10.70 ± 0.73 9±2 8.6 ± 1.2
Malt extract powder 4.48 ± 0.57 1.91 ± 0.57 3±1 1.6 ± 0.9

a
Analyzer output uncertainties are ±1s. Calculated uncertainties are ±2s.

where xm = measured mass fraction, and xc = known mass
fraction.
Uncertainties for corrected results include the average
correction factor standard deviation (SD) and the uncertainty
of the analyzer output.
Second, quadratic fits of the Br results for SRMs 1566,
1567, 1568, and 1570a (the first 4 listed in Table 1) were used
to calculate the system blank and produce a calibration curve.
Uncertainties for corrected results include those for the fit and
analyzer output.
Accuracy, Limits of Detection, and Interferences
When comparison values were available, XRF accuracy
was tested by using z-scores (20):
z = (xm – xc)/s
where xm = measured mass fraction, xc = known mass
fraction, and
s = s 2m + s 2c
where sm = 1s uncertainty of measured mass fraction, and sc
= 1s uncertainty of known mass fraction.
Z-score absolute values £2.0 indicate agreement, over 2.0
to £3.0 indicate questionable agreement, and >3.0 indicate
disagreement between results.
In previous work (9), raw spectral (channel-by-channel)
data for standards and deionized water were used to calculate
3s limits of detection (LODs), defined by Currie (21) as:
LOD = [2.71 + 4.65 ´ (continuum counts)1/2]/
[(count time) ´ (element sensitivity)]
In that study, analyzer LODs were found to correspond
closely to the calculated LODs. In this work, analyzer LODs
were corrected in the same way as were other results.
One key feature of the analyzer is that each X-ray spectrum
may be inspected on the iPAQ display. To check for
false-positive results, spectra need to be checked for possible
X-ray overlap caused by the presence of elements other than
that of the one of interest, in this case, Br. The Br Ka
(11.91 keV) X-ray may overlap the Hg Lb (11.87 keV) or As
Kb (11.73 keV) X-rays if those elements are present, and the
Br Kb X-ray (13.30 keV) may have an interference caused by
the Rb Ka (13.38 keV) X-ray.
Results and Discussion
Calibration with a Correction Factor
Based on % True Value Recovery
Analyzer results for the RMs are given in Table 1,
column 2. The %TVR values were calculated for 16 of the
RMs by comparing them to INAA and consensus values, and
plotting versus known Br mass fractions (Figure 1). By using
only %TVR data for Br >5 mg/kg, an average correction
factor (0.67 ± 0.11, 1 SD) was calculated to produce
reasonably accurate screening results (Table 1, column 4). The
large uncertainty (17%) associated with the factor precludes
use of this procedure for quantitative analysis. The correction
factor was applied to the analyzer results (corresponding to
rounded values of the data in Table 1, column 2) displayed on
the iPAQ to simulate a rapid screening situation. For single
portion analysis of RMs with Br mass fractions >5 mg/kg,
absolute values for z (not shown) were all <1.1 and were <0.9
for multiple portion analyses. Absolute z-scores for 2 low-Br
mass fraction SRMs (1568 Rice Flour and 1515 Apple
Leaves) were <2.0. The correction factor can be consistently
 ANDERSON: JOURNAL OF AOAC INTERNATIONAL VOL. 92, NO. 2, 2009 507
Figure 1. Known Br mass fractions for reference
materials compared to percent true value recoveries
and the correction factor determined for mass fractions
>5 mg/kg. Uncertainties are ±2s.
applied, even though portion densities varied significantly
(from about 0.3 to 1.0 g/mL), indicating that the instrument’s
built-in calibration with Compton normalization performs
adequately. The result for SRM 2711 (Montana Soil, a
nonorganic matrix) compared well to the NIST
information value.
The LOD of 9 mg/kg Br for SRM 2710 was high because
of the presence of strong Hg Lb X-rays at the same energy
(11.9 keV) as the Br Ka line used for the mass fraction
computation (Hg is certified at 32.6 ± 1.8 mg/kg for
SRM 2710). Also analyzed for comparison to quantitative
XRF results (see below), but not shown in Table 1 (because no
comparative values were available) were SRM 2709 San
Joaquin Soil (3 ± 2 mg/kg Br) and SRM 1566b Oyster Tissue
(53 ± 9 mg/kg Br). The LODs for SiO2 powder and deionized
water were calculated to be about 1 mg/kg Br. Even though
uncertainties were high (up to 25%, 2s) for RMs with Br
<5 mg/kg, reliable Br screening can be quickly performed for
a wide variety of powder matrixes.
The significant increase in %TVR observed for RM Br
mass fractions below about 5 mg/kg suggests the existence of
a system blank, possibly due to the presence of Br in the Mylar
film covering the test portion cell and Kapton window, or to
the nature of the factory-set calibration algorithm for this
element. Spectra for the SiO2 and deionized water blanks
showed no evidence of a Br Ka X-ray peak.
Calibration with NIST SRMs
To obtain more accurate results with smaller uncertainties,
the more precise readings (Table 1, column 2) given in result
files transferred from the iPAQ to PC were used for a more
detailed calibration. NIST SRMs 1568 Rice Flour, 1567
Wheat Flour, 1570a Spinach, and 1566 Oyster Tissue were
selected for use as standards because their mass fractions
represented the entire range of RM mass fractions fairly well.
Spectra for the 4 SRMs in the energy region of 10–14 keV are
shown in Figure 2. The As Ka X-ray at 10.5 keV in the SRM
1566 spectrum (As certified value = 13.4 ± 1.9 mg/kg) poses
no overlap problem for the calibration. As Kb X-rays at 11.7
interfere with the Br Ka X-ray for this SRM but constitute
<2% of the peak area. With the instrument calibration
algorithm correcting for the X-ray overlap, the interference is
negligible.
Analyzer output Br values (ppm) for the SRMs are shown
plotted versus INAA Br values (mg/kg) in Figure 3. A
nonweighted linear fit of the data yielded a y-intercept of 1.63 ±
1.06 ppm (R2 = 0.9989). Weighted linear fits yielded
y-intercepts of 2.44 ± 0.20 and 2.10 ± 0.88 ppm for INAA and
analyzer uncertainties, respectively, with R2 the same as for the
nonweighted fit. The average intercept (i.e., blank) value was
2.06 ± 0.41 (1 SD) ppm. A more consistent blank was obtained
by using quadratic fits to the data. Blanks obtained from
nonweighted, INAA uncertainty-weighted, and analyzer
uncertainty-weighted quadratic fits (all with R2 = 1.000) were
2.61 ± 0.12, 2.56 ± 1.02, and 2.54 ± 0.22 ppm, respectively,
with an average of 2.57 ± 0.04 (1 SD) ppm. For a calibration
curve, the INAA-determined mass fractions were plotted
versus the blank-corrected output data (Table 1, column 3) and
a nonweighted quadratic fit (R2 = 1.000) performed (Figure 4).
The linear fit (R2 = 0.9989) is shown for comparison.
Results (Table 1, column 5) for all RMs obtained by using
the calibration correction agree well with NIST Certificate
information values as well as with INAA and consensus
Figure 2. X-ray spectra over the energy range of
10–14 keV for NIST SRMs 1566, 1567, 1568, and 1570a
used for Br calibration.
508 ANDERSON: JOURNAL OF AOAC INTERNATIONAL VOL. 92, NO. 2, 2009
Figure 3. Nonweighted linear and quadratic fits of
INAA Br mass fractions for NIST SRMs 1566, 1567, 1568,
and 1570a versus blank-corrected analyzer output
(ppm). Data point uncertainties are ± 2s. For the linear
fit, M0 = 0.71 ± 0.75, M1 = 0.722 ± 0.017, R2 = 0.9989. For
the quadratic fit, M0 = 0.027 ± 0.012, M1 = 0.819 ± 0.001,
M2 = –0.00128 ± 0.00001, R2 = 1.000.
values. Uncertainties are generally smaller than those
calculated by using the %TVR correction factor, and the
results for the 2 correction methods agreed well. For SRM
1566 Oyster Tissue, SRM 1571 Orchard Leaves, SRM 1572
Citrus Leaves, RM 1577 Bovine Liver, and CFSAN Cocoa
Powder, multiple analyses of the same portion were
performed to check testing reproducibility. Here again,
z-scores indicated good agreement, with the possible
exception of SRM 1572 Citrus Leaves, for which there was
questionable agreement with the INAA and consensus values.
Results (not shown) for SRM 2709 and SRM 1566b were 1.2
± 2.1 and 55.1 ± 2.3 mg/kg Br, respectively, in agreement with
the correction factor values. The analyzer output value for the
SiO2 portion, 2.4 ppm (LOD), was consistent with the blank
level determined by quadratic fit, while the LOD for deionized
water was somewhat lower (1.4 ppm). Because no Br was
detected in these materials and SRM 2710, LODs were not
calculated.
Br in Foods
Results for TDS foods from MBs 2003-1, 2004-3, 2005-2,
and 2007-3 are presented in Table 2. With the exception of
MB 2007-3 foods, INAA comparison results are given. All Br
mass fractions were relatively low (ranging from “not
detected” to 6 mg/kg) compared to the ranges observed for
RMs. Relative uncertainties were large for both correction
methods. Agreement with INAA is reasonable, except for
food No. 178 chocolate cake, which includes nonflour icing.
INAA portions are typically smaller (0.3–1 g) than those for
XRF (2–7 g) and nonhomogeneity may play a role in the
disagreement. Consistent with Figure 1, use of the correction
factor yielded higher results for low-Br matrixes compared
with those calculated from the quadratic fit. Despite this, both
correction methods were accurate in that mass fractions below
regulatory levels could be confirmed in a wide variety of
matrixes. A simple correction factor can be used for screening
purposes, as confirmed by the agreement with the results from
the empirical calibration approach.
Br in Flour and Grain Products
Results for commercial brands of 2 bleached flours,
2 breads, 1 potato roll, 2 cookies, 3 crackers, and
3 grain-based snack foods are presented in Table 3. The
spectrum obtained for bleached flour (brand No. 2, 6 mg/kg
Br) is compared with that of SRM 1570a (35.5 mg/kg Br) in
Figure 5. This spectrum is typical of all the foods tested in this
work, i.e., no measurable interferences from As, Hg, or Rb
were observed. Br levels measured in these products ranged
from below detection to 13 mg/kg, well below regulatory
limits. The 2 primarily nonwheat-based snack foods had
nondetectable levels of Br. Results from the 2 correction
methods agreed very well at Br mass fractions >5 mg/kg. Here
also, use of the correction factor yielded higher results for
low-Br foods compared with those calculated from the
quadratic fit. Analysis of one brand of flour through a
polyethylene “baggie” yielded results about 30% higher than
the X-ray cell analysis, and bread analyzed the same way
yielded results about 80% higher. Similar increases were
observed for the 2 breads and 1 cookie analyzed directly,
Figure 4. Nonweighted linear and quadratic fits of
analyzer output (ppm) versus INAA Br mass fractions
for NIST SRMs 1566, 1567, 1568, and 1570a. Data point
uncertainties are ±2s. For the linear fit, M0 = 1.63 ± 1.06,
M1 = 1.38 ± 0.03, R2 = 0.9989. For the quadratic fit,
M0 = 2.61 ± 0.12, M1 = 1.194 ± 0.013, M2 = 0.00348 ±
0.00024, R2 = 1.000.
 ANDERSON: JOURNAL OF AOAC INTERNATIONAL VOL. 92, NO. 2, 2009 509
Figure 5. X-ray spectra over the energy range of
10–14 keV for bleached flour and SRM 1570a Spinach.
implying that the X-ray cells’ Mylar cover can affect detection
efficiency because of X-ray absorption.
Br in Malted Barley Products
Br results for malted barley products are shown in Table 4.
The whole grain malts were first analyzed as received, then
reanalyzed after pulverization. Visibly, the malted barley
powders matched the RM powder matrixes more closely than
the whole grains. The whole grain malts toasted wet and dry
had been prepared from the plain malt. The malt extract
powder closely matched the CFSAN Cocoa Powder matrix
because of the high sugar content, but because the malt extract
syrup is a liquid matrix, the calibration may not provide an
accurate measurement. Whole grain and powder results
agreed within the uncertainties except for toasted dry malt, for
which the whole grain result was almost twice that for the
powder. Nonhomogeneity is a possible explanation, but the
discrepancy was not shown in the original plain malt or the
toasted wet malt. Here again, the factor-corrected results
agreed well with those from the quadratic fit, and all Br levels
found were well below regulatory limits.
Conclusions
Flour, bakery products, and malted barley products can be
screened quickly for Br at levels >5 mg/kg with a
field-portable X-ray tube analyzer. A simple correction factor
can be applied to the analyzer output to achieve accurate
results, albeit with rather large (up to 25%) relative
uncertainties. Quantitative results with smaller uncertainties
can be obtained for matrixes of similar composition by using a
quadratic fit of known Br values versus blank-corrected
analyzer output. The quantitative results verified accuracy of
those obtained by using the simple correction factor.
Preparation of homogeneous matrix-matching (e.g.,
cellulose-based or spiked food) standards could decrease
uncertainties, improve accuracy, and extend the calibration
range upward to include the higher regulatory limit of
75 mg/kg Br. The 2 min analysis times were chosen as an
approximation of the maximum time feasible for repetitive
handheld analyses, because of fatigue considerations. With
use of the testing stand, analysis times could be increased,
which would significantly lower LODs and uncertainties. For
products with Br levels near regulatory limits of 24 mg/kg
(flour), 36 mg/kg (whole wheat flour and bakery products),
and 75 mg/kg (barley malt), shorter analysis times
(30 s–1 min) would suffice for accurate measurements.
Although no calibration was performed in this study for Br in
liquids, a previous report (22) showed that rapid screening and
quantitation were possible for As, Hg, and Pb in beverages at
similar mass fraction. Thus, it is likely that malt beverages
(regulatory limit of 25 mg/kg Br) could likewise be
successfully analyzed for Br by using the field-portable
analyzer.
Acknowledgments
I thank the Operations Staff of the NIST Center for
Neutron Research for their constant support, Richard Jacobs,
FDA, for fruitful discussions on XRF spectrometry, Bill
Cunningham, FDA, for SRM Br data, and Mark English,
NIST, for supplying the malted barley products.
References
(1) Kurokawa, Y., Hayashi, Y., Maekawa, A., Takahashi, M.,
Kokubo, T., & Odashima, S. (1983) J. Nat. Cancer Inst. 71,
965–971
(2) International Agency for Research on Cancer, World Health
Organization (1986) Monographs on the Evaluation of the
Carcinogenic Risk of Chemicals to Humans, Some Naturally
Occurring and Synthetic Food Components, Furocoumarins,
and Ultraviolet Radiation, IARC/WHO, Vol. 40, Lyon,
France, pp 207–220
(3) Cunningham, W.C., & Warner, C.R. (2000) Food Addit.
Contam. 17, 143–148
(4) Briggs, D.E., Hough, J.S., Stevens, R., & Young, T.W. (1981)
Malting and Brewery Science, Kluwer Academic/Plenum
Publishers, New York, NY
(5) U.S. Code of Federal Regulations (1994) 137.155, 137.160,
137.205, http://www.gpoaccess.gov/CFR/INDEX.HTML
(6) U.S. Code of Federal Regulations (2003) 136.110, 136.115,
136.130, 136.160, 136.180), http://www.gpoaccess.gov/
CFR/INDEX.HTML
(7) U.S. Code of Federal Regulations (2007) 172.730,
http://www.gpoaccess.gov/CFR/INDEX.HTML
(8) Anderson, D.L. (2003) J. AOAC Int. 86, 583–597
(9) Anderson, D.L. (October 24, 2006) Detection of Toxic
Elements in Foods With a Portable X-Ray Tube Hand-held
Analyzer, Research Report, U.S. Food and Drug
Administration, Center for Food Safety and Applied
Nutrition, College Park, MD
(10) Palmer, P., Webber, S., Ferguson, K., & Jacobs, R. (2006) On
the Suitability of Portable X-ray Fluorescence Analyzers for
Rapid Screening of Toxic Elements, Laboratory Information
510 ANDERSON: JOURNAL OF AOAC INTERNATIONAL VOL. 92, NO. 2, 2009
Bulletin 4376, U.S. Food and Drug Administration, Division
of Field Science, Rockville, MD
(11) U.S. Food and Drug Administration (2009) Total Diet Study,
http://www.cfsan.fda.gov/~comm/tds-toc.html
(12) Anderson, D.L., & Cunningham, W.C. (2000) J. AOAC Int.
83, 1121–1134
(13) Anderson, D.L., & Cunningham, W.C. (2008) J. Radioanal.
Nucl. Chem. 276, 23–28
(14) Bertin, E.P. (1975) Principles and Practice of X-ray
Spectrometric Analysis, Plenum Press, New York, NY
(15) Jordan Valley, Austin, TX and Analytical Reference
Materials International, Golden, CO,
http://www.learnxrf.com/
(16) Innov-X Systems, Inc., Woburn, MA,
http://www.innovxsys.com
(17) Anderson, D.L. (2006) Detection of Toxic Elements in Foods
with a Portable X-Ray Tube Analyzer, Elemental Research
Branch Standard Operating Procedure No. 114, U.S. Food
and Drug Administration, College Park, MD
(18) Cunningham, W.C., Capar, S.G., & Anderson, D.L. (1997)
J. AOAC Int. 80, 871–882
(19) Roelandts, I., & Gladney, E.S. (1998) Fresenius Z. Anal.
Chem. 83, 327–338
(20) Thompson, M., & Wood, R. (1993) J. AOAC Int. 76,
926–940
(21) Currie, L.A. (1968) Anal. Chem. 40, 586–593
(22) Anderson, D.L. (2008) Analysis of Beverages for Hg, As, Pb,
and Cd with a Field-Portable XRF Analyzer, Research
Report, U.S. Food and Drug Administration, College Park,
MD