Determination of boron in materials by cold neutron prompt gamma-ray activation analysis

Abstract

An instrument for cold neutron prompt gamma-ray activation analysis (PGAA), located at the NIST Center for Neutron Research (NCNR), has proven useful for the measurement of boron in a variety of materials. Neutrons, moderated by passage through liquid hydrogen at 20 K, pass through a ⁵⁸Ni coated guide to the PGAA station in the cold neutron guide hall of the NCNR. The thermal equivalent neutron fluence rate at the sample position is 9 × 10⁸ cm⁻² s⁻¹. Prompt gamma rays are measured by a cadmium- and lead-shielded high-purity germanium detector. The instrument has been used to measure boron mass fractions in minerals, in NIST SRM 2175 (Refractory Alloy MP-35-N) for certification of boron, and most recently in semiconductor-grade silicon. The limit of detection for boron in many materials is <10 ng g⁻¹.

---

Introduction

The determination of boron in materials is important to many fields of science. Boron is added to steels at low levels (≈0.003%) to strengthen grain boundaries, increase hardness, and improve mechanical properties. It is doped into semiconductor materials to alter electrical properties, and is the most important p-type dopant in silicon. Boron is also an important dietary element.

Prompt gamma-ray activation analysis (PGAA) is particularly sensitive to the measurement of boron mass fractions, with detection limits at ng g⁻¹ levels. Although boron may also be determined by inductively coupled plasma mass spectrometry (ICP-MS) and inductively coupled plasma optical emission spectrometry (ICP-OES), PGAA has the advantage of being nondestructive, allowing boron to be analyzed in situ. Therefore the losses and contamination during chemical processing that are inherent in other techniques may be avoided. There is a theoretical disadvantage in that only ¹⁰B is measured.

Thermal neutron PGAA has been used at NIST since 1979 to determine boron in reference materials.^1–4 An instrument for cold neutron PGAA, operational in the cold neutron guide hall of the NIST Center for Neutron Research (NCNR) since 1991, has the advantage of lower detection limits due to higher neutron capture rates and boron-free construction. This instrument is part of a national users’ facility and is available to scientists worldwide through submission of proposals The measurement of boron content of various materials by this instrument is the subject of this study.

Experimental

Apparatus

The cold neutron prompt gamma-ray spectrometer has been described in detail previously.^5–7 Neutrons, moderated by passage through liquid hydrogen at 20 K, pass through a ⁵⁸Ni coated guide to the PGAA station. The thermal equivalent neutron fluence rate at the sample position is 9 × 10⁸ cm⁻² s⁻¹. Prompt gamma rays emitted by the sample upon neutron capture are measured by a cadmium and lead shielded high-purity germanium detector (35% efficiency, 1.7 keV resolution). The germanium detector is positioned vertically inside a bismuth germanate (BGO) Compton suppression detector, which reduces the spectrum baseline (increasing signal-to-noise ratio) by gating out signals due to Compton scattered gamma rays (detected by both detectors simultaneously). Signal processing is accomplished by means of standard Nuclear Instrument Modules (NIMs)†. Gamma-ray spectra up to 11 MeV are collected using a multichannel analyzer that is controlled using a VAXstation.

Sample preparation and irradiation

The sample preparation procedure was dependent upon the type of material analyzed. Mineral separates were analyzed for the purpose of developing standards for geochemical analysis. The minerals analyzed included sillimanites (Al₂O₃SiO₂), muscovites (K₂O·3Al₂O₃·6SiO₂·2H₂O), and biotites K₂(Mg,Fe)₃AlSi₃O₁₀(OH,O,F₂)₂. Mineral separates were first crushed using an agate mortar, then weighed into a mixing vial along with a weighed quantity of graphite (approximately 40–70 mg of mineral mixed with 300 mg of graphite). Mixing was performed using a Spex Mixer Mill. Each mixture was then pressed into one or more pellets (each ≈ 300 mg) using a 12.7 mm diameter stainless steel die and a hydraulic press. Pellets were sealed into bags of FEP Teflon^® (hereafter referred to as Teflon) for PGAA analysis.

PGAA was used to assign the boron mass fraction value in NIST SRM 2175 (Refractory Alloy MP-35-N) for the purpose of certification. The metal, an alloy of cobalt, nickel, chromium, and molybdenum, was in the form of small chips. Replicate portions, ≈50 mg each, were weighed into Teflon bags and sealed.

Boron-10 was determined at native levels in silicon as part of a CCQM (Comité Consultatif pour la Quantité de la Matière) study. The silicon wafer production industry commonly uses infrared (IR) spectroscopy, photoluminescence, and electrical resistivity methods to measure the boron content of silicon. However, these are not primary methods, and calibration factors are necessary to calculate the boron content of the silicon. The CCQM therefore commissioned the following pilot study in order to determine whether primary methods could be used to determine the mass fraction of boron in silicon. An unmodified 200 mm wafer material, designated CCQM-P33, supplied by a German silicon manufacturer was used in this study. The boron content, according to electrical resistivity measurements, was specified to be ω(B) ≈ 40 ng g⁻¹; preliminary studies using ICP-MS yielded a boron mass fraction of 70 ng g⁻¹. Samples (≈1 cm × 1 cm), obtained by breaking up one wafer, were distributed to participants in the study. Eight of these samples were measured by cold neutron PGAA. Samples for PGAA were cleaned by soaking in ethanol and ultrapure water, and were then allowed to dry. Each sample was then sealed into a Teflon bag.

All samples were mounted by suspension between Teflon strings and irradiated at the cold neutron PGAA station. Irradiation times, dependent on the sample size and amount of boron in the sample, varied from 2 h for the alloy samples to up to 2 d for the silicon. Changes in neutron fluence rate were monitored by periodic irradiation of a titanium foil. Over the course of a day, the Ti count rate variation was generally <0.5%.

Standards

For measurement of the boron content of minerals, a set of standards was prepared to yield a matrix similar to that of the samples. A solution containing 0.5938 mg mL⁻¹ boron was prepared by dissolving boric acid (NIST standard reference material 951, boron mass fraction certified at 17.586%) in ultrapure water. A 50.0 µL aliquot of this solution was deposited onto silicon dioxide which, after drying, was mixed with graphite in a Spex mixer mill. Boron standards were prepared by pressing the mixture into pellets. Each pellet contained ≈9 µg of boron.

In order to minimize the effects of neutron self shielding in the analysis of the refractory alloy SRM, the ratio of B to Co in the metal was measured.⁸ The boron standards prepared above were used in this analysis, along with a 46 mg Co foil (σ_capture = 37 b). Additional standards for ratio measurement were prepared by mixing 0.9979 mL of cobalt solution (NIST spectrometric solution SRM 3113, 10.00 mg mL⁻¹) with 0.0514 mL of the above mentioned boron solution and sealing the mixtures in Teflon bags.

NIST Standard Reference Material 2137 (Boron in Silicon Implant), certified at (0.016 92 ± 0.000 89) µg ¹⁰B cm⁻², was used as a standard for the CCQM silicon samples. A 1 cm × 1 cm portion of the SRM was cleaned in the same manner as the silicon samples and sealed into a Teflon bag.

Background correction

In the absence of a sample in the beam, a small boron count rate is detected, presumably due to neutron capture by boron in an adjacent neutron guide. This boron background count rate was measured by irradiating an empty Teflon bag for at least 12 h. For analysis of mineral samples, blank measurements were also made by irradiating pellets of graphite and of graphite mixed with SiO₂. All measurements yielded boron count rates of the order of 0.1 counts s⁻¹. This background was negligible for measurement of boron in the alloy and most of the mineral separates.

The boron background was not negligible in the analysis of the CCQM silicon, the background comprising about 10% of the total count rate. A major concern with these measurements was that the reagents used to clean the silicon samples and Teflon bags might actually introduce boron contamination. To settle this issue, two background measurements were made, one by irradiation of a Teflon bag without washing, and a second background by irradiation of a Teflon bag that had been washed and dried in the same manner as the samples. The boron count rates for the two bags were identical within counting statistics, indicating that washing did not add or remove a significant amount of boron. Thus we conclude that the irradiated bag represents a true background (blank) and that boron is not introduced by cleaning.

Nuclide interferences

The net peak area for the Doppler-broadened 477 keV photopeak resulting from the ¹⁰B(n, α)⁷Li reaction (Fig. 1) was determined by using the SUM4 code.⁹ Corrections for interference peaks were made by measuring the count rate ratio of the interference peak to a reference peak in the prompt gamma-ray spectrum of a boron-free standard of the interfering element (Table 1). This reference peak was then measured in the sample spectrum, and the interference correction calculated by multiplying the reference peak count rate by the interference/reference peak ratio determined from the standard. It was important to choose a reference peak that was free of interferences in the sample spectrum.

Fig. 1 Boron peak measured in a boron standard (H₃BO₃/SiO₂/graphite pellet).

Table 1 Significant nuclide interference peaks for boron noted in the samples analyzed

Element	Interference peak/keV	Reference peak(s)/keV^a	Ratio (interference peak/reference peak)^a
a The reference peaks used in the calculation of correction factors are also shown.
Na	472.2	874.3	9.4
Mo	480.6, 480.9	778	0.046
Ni	483.4	465	0.017
Co	484.3	230, 277, 556	0.088 (230), 0.095 (277), 0.16 (556)
Si	477	752.2	0.044

---

Boron count rates for the mineral separates were corrected for a sodium interference at 472.2 keV (Fig. 2). This correction was relatively small for these materials, ≤5% of the total count rate for all but one of the samples measured.

Fig. 2 Boron peak measured in a muscovite sample, with sodium interference.

Boron count rates for the alloy samples were corrected for a large cobalt interference at 474.3 keV and smaller interferences from nickel (483.3 keV) and molybdenum (480.5 keV, 480.9 keV) (Fig. 3). The cobalt correction was about 50% of the count rate of the combined peak.

Fig. 3 Boron peak measured in a sample of SRM 2175 (Refractory Alloy MP-35-N) with cobalt interference at 484 keV.

For measurement of boron at nanogram levels in CCQM silicon, a silicon interference at 477 keV yielded a count rate equivalent to about 50% of the combined peak (Fig. 4). The correction factor was smaller for the SRM 2137 standard because of the higher boron-to-silicon ratio.

Fig. 4 (a) Boron peak measured in a sample of CCQM silicon showing silicon interference. (b) Boron peak measured in SRM 2137 (Boron in Silicon Implant).

Results and discussion

The results of the various analyses are given below. Unless otherwise specified, boron mass fractions were determined assuming identical isotopic compositions for the boron in the samples and standards. Measurement uncertainties were evaluated using guidelines set by the International Organization for Standardization (ISO).¹⁰

Table 2 gives boron mass fractions determined for mineral separates. Mass fractions range from to 2 mg kg⁻¹ to 200 mg kg⁻¹. The boron mass fraction determined by PGAA for the control material, SRM 1633a (Coal Flyash), is in agreement with a consensus value reported by Gladney et al.¹¹ Only one other independent boron measurement has been made on any of these samples (Kerrick's sillimanite by atomic absorption spectrometry (AAS)), and this measurement is in agreement with the PGAA value. Secondary ion mass spectroscopy (SIMS) measurements have been made on the sillimanite and muscovite samples, using the PGAA data for calibration (Fig. 5).

Fig. 5 Calibration plot for measurement of boron in minerals by SIMS, using muscovite standards calibrated by PGAA. The broken line shows the linear calibration calculated only from the data. The continuous line, obtained by forcing the plot to pass through the origin, is the actual calibration used.

Table 2 Boron mass fractions measured in mineral separates and in control material SRM 1633a (Coal Flyash)

Sample	Boron mass fraction/mg kg^−1
	PGAA^c	Other measurements
a Atomic absorption measurement, J. Husler, University of New Mexico. (Private communication from Edward Grew, University of Maine.) b Consensus value from E. S. Gladney et al., ref. 10. c Expanded uncertainties (k = 2) were evaluated using uncertainties arising from counting statistics, irradiation geometry and variations in neutron fluence rate, sample inhomogeneity due to mixing, and preparation of standards.
556 (sil)	19.1 ± 1.2
H-944 (sil)	2.5 ± 0.2
AD13 (sil)	21.7 ± 1.4
Kerrick's (sil)	200 ± 7	200^a
2–8/7/63 (musc)	45.0 ± 2.7
22–8/7/58 (musc)	43.8 ± 2.6
6–8/7/63 (musc)	48.3 ± 2.9
17–8/14/59 (musc)	63.4 ± 3.7
7–8/5/59 (musc)	29.3 ± 1.8
7–8/3/59 (musc)	32.8 ± 2.1
27–9/10/59 (musc)	19.8 ± 1.2
5–9/20/61 (bio)	11.1 ± 0.7
16–7/18/60 (bio)	17.1 ± 1.0
7–8/5/59 (bio)	2.0 ± 0.2
Control material—SRM 1633a, Coal Flyash	39.4 ± 2.1	40.3 ± 2.1^b

---

Table 3 gives boron mass fractions determined for SRM 2175 (Refractory Alloy MP-35-N). In order to minimize uncertainties due to neutron self shielding, the ratio of the boron mass fraction to the cobalt mass fraction was determined from the ratio of the boron/cobalt count rates measured in the sample and boron and cobalt sensitivities measured in a mixed standard. The boron mass fraction for each sample was then determined by normalizing this ratio to a cobalt mass fraction consensus value of 33.3% measured by ICP-OES, instrumental neutron activation analysis (INAA), and X-ray fluorescence (XRF). For the purpose of quality control, the boron/cobalt mass fraction ratio was also determined for SRM 349a (Waspaloy) with certified mass fractions of (0.005 ± 0.001)% for boron and (12.46 ± 0.08)% for cobalt. The ratio measured by PGAA was in agreement with the ratio of mass fractions calculated from the boron and cobalt certified values. The mean boron mass fraction of (95.2 ± 4.4) mg kg⁻¹ determined for SRM 2175 was combined with measurements from outside laboratories to yield the certified value of (97 ± 23) mg kg⁻¹.

Table 3 Results of analysis of SRM 2175, Refractory Alloy

Sample	(B mass fraction)/(Co mass fraction)		B mass fraction/mg kg^−1
	PGAA^a	From certified values^a
a The expanded uncertainty (k = 2) for the final value was evaluated using uncertainties arising from measurement replication, determination of interference corrections, normalization to cobalt, gamma-ray attenuation, and preparation of standards.
SRM 2175 9	0.000 278		93.5
SRM 2175 178	0.000 281		94.6
SRM 2175 268	0.000 279		93.8
SRM 2175 416	0.000 283		94.9
SRM 2175 442	0.000 290		97.6
SRM 2175 449	0.000 287		96.6
Average value	0.000 283		95.2
Final value			95.2 ± 4.4
Control material—SRM 349a (Waspaloy)	0.000 399 ± 0.000 020	0.000 401 ± 0.000 080

---

Table 4 gives boron-10 mass fractions determined in the CCQM silicon samples. The ¹⁰B/Si ratio was measured in order to minimize uncertainties. This ratio was determined by measuring the ratio of count rates of the boron 478 keV and silicon 1273 keV peaks in the samples and comparison of boron and silicon sensitivities determined from the SRM 2137 standard. A complete evaluation of uncertainties for ¹⁰B is given in Table 5. The mean ¹⁰B mass fraction was found to be (10.5 ± 1.7) × 10⁻⁹ g g⁻¹ (expanded uncertainty). Assuming a natural atomic abundance of 0.199 for ¹⁰B, this corresponds to a mean mass fraction for total boron of 53 × 10⁻⁹ g g⁻¹ . However, since ICP-MS measurements of the silicon indicate some deviation from normal isotopic composition, this value for total boron should be considered an approximation, and is listed only to give the reader a feel for the total boron content of the wafer.

Table 4 ¹⁰B mass fractions measured in silicon (CCQM-P33)

Sample	^10B (×10^−9 g g^−1
N02	10.0
N03	9.68
N04	11.5
N05	9.73
N06	10.8
N10	10.3
N12	11.0
N18	11.1
Average value	10.5
Standard deviation	0.7

---

Table 5 Uncertainty budget for determination of ¹⁰B in silicon (CCQM-P33)

Source of uncertainty	Uncertainty in ^10B mass fraction × 10^−9/g g^−1
Measurement replication (s/√n)	0.24
Interference correction	0.1
Background subtraction	0.7
Standard: counting statistics	0.32
Standard concentration	0.17
Combined uncertainty	0.84
Coverage factor	2
Expanded uncertainty	1.7

---

Limits of detection

Limits of detection (LOD) for boron in selected matrices by cold neutron PGAA have been calculated using an equation derived from the works of Jaklevic and Walter¹² and Currie:¹³

LOD = 4.65 (R_b/t)^1/2/S

where R_b is the background counting rate (counts s⁻¹), t is the duration of the count (s), and S is the hydrogen sensitivity in counts s⁻¹ mg⁻¹. Calculated for a 1 g sample irradiated for 24 h, boron limits of detection are 6 ng g⁻¹ in silicon, 13 ng g⁻¹ in Coal Flyash (SRM 1633a), 6 ng g⁻¹ in Pine Needles (SRM 1575a) and 14 ng g⁻¹ in stainless steel.

Conclusions and future work

Cold neutron PGAA has proven useful for the determination of low-level boron in a variety of materials. In the future, the applicability of the technique will be enhanced by applying existing technologies and planned upgrades to the instrument. The application of neutron focusing techniques^14–16 (which can focus the beam down to a spot size of 0.5 mm, resulting in a factor of ≈20 gain in neutron flux over the unfocused beam) has made it possible to probe the boron content of materials as a function of position. Future plans to bend the PGAA neutron beam away neutron guide NG7 will make it possible to analyze larger samples, giving us the capability to analyze an entire 8 in silicon wafer for boron as a function of position. This upgrade may also result in reduction of the boron background from neutron capture in the adjacent guide.

Acknowledgements

The author wishes to thank Dr. Edward Grew of the University of Maine for providing mineral samples for analysis as well as SIMS calibration data, and Dr. Reinhard Jährling of the Physikalisch-Technische Bundesanstalt, Germany, for supplying the CCQM silicon samples. The author also wishes to thank the staff of the NCNR.

References

1. M. P. Failey, D. L. Anderson, W. H. Zoller, G. E. Gordon and R. M. Lindstrom, Anal. Chem., 1979, 51, 2209–2221.
2. D. L. Anderson, W. C. Cunningham and E. A. Mackey, Fresenius’ J. Anal. Chem., 1990, 338, 554–558.
3. M. Darby Dyar, M. Wiedenbeck, D. Robertson, L. R. Cross, J. S. Delaney, K. Ferguson, C. A. Francis, E. S. Grew, C. V. Guidotti, R. L. Hervig, J. M. Hughes, J. Husler, W. Leeman, A. V. McGuire, D. Rhede, H. Rothe, R. L. Paul, I. Richards and M. Yates, Geostand. Newsl., 2001, 25(2–3), 441–463.
4. D. L. Anderson, Y. Sun, M. P. Failey and W. Zoller, Geostand. Newsl., 1985, 9, 219–228.
5. R. L. Paul, R. M. Lindstrom and A. E. Heald, J. Radioanal. Nucl. Chem., 1997, 215(1), 63–68.
6. R. L. Paul and R. M. Lindstrom, J. Radioanal. Nucl. Chem, 2000, 243, 181–189.
7. R. L. Paul, Am. Lab., 2002, 34(3), 15–20.
8. R. L. Paul, J. Radioanal. Nucl. Chem., 1995, 191(2), 245–256.
9. R. M. Lindstrom, Biol. Trace Elem. Res., 1994, 43–45, 597–603.
10. Guide to the Expression of Uncertainty in Measurement, ISO, Switzerland, 1st edn., 1993, ISBN 92-67-10188-9.
11. E. S. Gladney, B. T. O'Malley, I. Roelandts and T. E. Gills, Standard Reference Materials: Compilation of Elemental Concentration Data for NBS Clinical, Biological, Geological, and Environmental Standard Reference Materials, NBS Special Publication 260-111, U.S. Department of Commerce, 1987.
12. J. M. Walter and R. L. Walter in X-Ray Fluorescence Analysis of Environmental Samples, ed. T. G. Dzubay, Ann Arbor Science, Ann Arbor, MI, 1977, ch. 5.
13. L. A. Currie in X-Ray Fluorescence Analysis of Environmental Samples, ed., T. G. Dzubay, Ann Arbor Science, Ann Arbor, MI, 1977, ch. 25.
14. H. Chen, V. A. Sharov, D. F. R. Mildner, R. G. Downing, R. L. Paul, R. M. Lindstrom, C. J. Zeissler and Q.-F. Xiao, Nucl. Instrum Meth. B, 1995, 95, 107–114.
15. H. H. Chen-Mayer, E. A. Mackey, R. L. Paul and D. Mildner, J. Radioanal. Nucl. Chem, 2000, 244(2), 391–397.
16. H. H. Chen-Mayer, W. J. Heward, R. L. Paul, F. J. Klug and Y. Gao, J. Mater. Res., 2003, 18(10), 2486–2493.

---

Footnote

† The identification of certain commercial equipment, instruments, or materials does not imply recommendation or endorsement by the National Institute of Standards and Technology. These identifications are made only in order to specify the experimental procedures in adequate detail.

---