Elemental analysis of concrete with fast neutron beams

Within nuclear installations concrete is widely used for its structural and radiation shielding properties. Developing methods to independently, and non-destructively, verify the composition of such materials in situ is of high priority for the regulation of existing and future nuclear installations. This work reports on the ongoing development of measurement techniques utilising fast neutron beams for the elemental analysis of concrete and its constituents.


Introduction
The non-destructive elemental analysis of bulk materials such as concrete, in situ, is of direct importance to the regulation of nuclear facilities.In modern installations, licencing procedures require knowledge of the bulk shielding composition at the design stage, however, for older facilities designed to earlier standards this information is not typically available [1].Ordinary concrete is a seemingly simple mixture of sand and other aggregates, cement and water.However, concrete often contains voids, inconsistent distribution of aggregates and variable moisture content depending on the preparation [2].In nuclear facilities concrete is often exposed to extreme conditions resulting in a time-varying composition, particularly with respect to the water content, which is a concern for ageing nuclear installations [3].The use of radiation transport simulations is widely used for the design and management of nuclear facilities, and poor knowledge of the concrete composition is known to cause significant discrepancies between measured and simulated data [4].
Neutron transmission analysis considers the change in the neutron fluence spectrum when a beam of neutrons is incident upon a sample.For a bulk material with thickness , the attenuation of a neutron beam with incident fluence  0 () is described by Eq. 1 where, () is the transmitted fluence [5], and the energy dependent effective removal cross section Σ  () is characteristic of each element [6].For a composite material with density , the effective removal cross section can be defined as a linear sum of its constituent parts , weighted by their mass ratios   as defined in Eq. 2.
For an unknown sample, the relative amounts of each element can be determined from the measured energy dependent removal cross section, provided the removal cross sections are known for each element of interest.In this work we present simulation-based proof-of-concept results for the case of sand, a major component of any concrete, and the first experimental verification of the technique, which builds upon previous investigations at the University of Cape Town [2,5].

Radiation transport simulations
Monte Carlo radiation transport simulations were undertaken with MCNP6 [7] to determine the energy dependent effective removal cross sections for H, C, O, Si and Ca.The neutron source was modelled as a 0.8 cm diameter pencil beam distributed in energy according to the ISO recommended 241 Am-9 Be (AmBe) spectrum between 1.0 -11.0 MeV [8].The sample was defined as a Ø 5.0 cm x 5.0 cm cylinder aligned with the beam axis, with densities chosen to approximate what may be physically obtainable in the laboratory (H 0.7 g cm -3 ; C 1.4 g cm -3 ; O 2.1 g cm -3 ; Si 2.3 g cm -3 ; Ca 1.5 g cm -3 ).A spherical detection volume (Ø 5.0 cm) was placed at 50.0 cm from the centre of the sample to score the energy dependent neutron fluence () (F4 tally).The energy dependent fluence was used to determine the effective removal cross section for each element according to Eq. 1, where  0 () is incident fluence.
A second set of simulations were undertaken for simple compounds (H2O, SiO2, CaCO3), and sand (2 % H2O; 65 % SiO2; 33 % CaCO3) to investigate the use of the MAXED unfolding code [9] to deconvolve the mass ratios (  ) where the truth value is known.Elemental unfolding requires the measured signature for the sample (Σ  ()  ⁄ , or some variation thereof), a series of elemental response functions (Σ , ()   ⁄ ), and an initial estimate of the mass ratios of each element.Poor knowledge of the sample composition is handled assuming a priori equal ratios of all elements.

Experimental validation
The elemental unfolding process was validated with experimental measurements undertaken at the n-lab [10].A pencil beam of neutrons (Ø 0.8 cm) was produced with a 220 GBq AmBe source and HDPE collimator.A 6.0 cm thick sample of SiO2 beads (< 0.1 cm diameter) was prepared with a crosssectional area of 5.0 x 5.0 cm 2 .Neutrons (and gamma rays) were detected using a 2" x 2" EJ301 organic liquid scintillator, and data were acquired digitally [11].The light output parameter L was taken to be the pulse height of the dynode signal after shaping and amplification and calibrated with a series of gamma ray sources.Neutron and gamma ray events were separated via the charge comparison method of pulse shape discrimination using the fast anode signal [12] to obtain the neutron-only light output spectra.Spectrum unfolding with MAXED, and a set of monoenergetic detector response functions, was then used to determine the neutron fluence spectra and subsequently the measured Σ  () using Eq. 1.

Radiation transport simulations
The simulated transmitted neutron fluence is shown in Fig. 1 for no sample, and the five elemental samples.In each instance, the total number of simulated histories was varied to achieve similar levels of convergence, and statistical uncertainties on the integral fluence below 0.1 %.These fluence spectra were used to calculate (Eq. 1) the elemental response functions, Σ , ()/  , and are shown in Fig. 2.  Transmitted neutron energy spectra were simulated for the composite samples and used to determine Σ  () as the sample signature.It was assumed that the density  of the material was unknown, which is often the case for measurements made in situ.The unfolded solution would therefore be equal to   for each element , from which the density was obtained.The expected and unfolded densities are shown in Table 1, and consistently show excellent agreement, which is useful for investigations into the homogeneity of a sample along the path of the beam.Table 1.Density reconstruction from elemental unfolding analyses.
The elemental composition of each sample was unfolded from the measured signatures, and the resulting mass fractions are shown in Fig. 3. Overall, the agreement between the expected and unfolded ratios is considered good.In particular, the H content in H2O is very well represented within uncertainties (11.1 % expected, 11.1 ± 0.9 % unfolded).For the case of sand, the mass ratio of H is low (0.21%), with an unfolded value of 0.31 ± 0.06 %.The slight discrepancy is attributed to the low probability of an interaction with H in the bulk material, which suggests that the sensitivity of the technique is limited to around 0.1 % by mass for H in these simulated data.Similar behaviour is seen in the CaCO3 data, where there are insufficient neutron interactions within the sample due to a low material density, resulting in inconsistencies (3 -5 %) in the C, Si and Ca content.

Experimental validation
Measurements were made of the transmitted neutron spectra for neutrons produced by an AmBe source through a 6.0 cm thick sample of SiO2.The neutron energy spectra shown in Fig. 4 were unfolded from the calibrated neutron light output spectra with MAXED, using the ISO recommended spectrum as the initial estimate of the energy distribution.A coarse binning structure (1 MeV) was chosen to minimise the effect of the measured statistics in L on the unfolding, and account for small deviations in gain between measurements.The measured signature for SiO2 was then obtained via the same process used for the simulated data and is shown in Fig. 5, along with the refolded signature.

Sample
(g cm   The overall shape of this signature is well matched to the simulated equivalent (Fig. 2), except for an enhancement in the measured signature at 8.0 MeV, which is attributed to a slight calibration offset between the no sample and SiO2 measurements.A subset of energies (1.0 -8.0 MeV) were used, in combination with the simulated elemental response functions, to unfold the elemental composition to be 1 % H; 6 % C; 47 % O; 41 % Si; and 5 % Ca.The expected composition for SiO2 is 53 % O and 47 % Si by mass.An unfolded composition within 6 % of the expected value at this stage is very promising, both for the techniques, and the use of simulated response functions to unfold measured data.

Conclusion
Fast neutron transmission spectroscopy, and the use of composition unfolding with elemental response functions, has been demonstrated to be a powerful technique in the analysis of bulk materials such as concrete.The use of radioisotopic neutron sources, digital data acquisition and unfolding allows the technique to be deployed for in situ analyses and is not restricted use in the laboratory.Preliminary investigations indicate that simulated elemental removal cross sections are applicable in the analysis of measured data, which is useful in cases where elemental media are not easily obtained.Further validation is ongoing across a range of elements and energies.The developed technique, in its present form, is most sensitive to the analysis of low Z elements due to the distinct energy dependent features present in the reaction cross sections.To increase the differentiability between higher Z constituents, the addition of complementary radiation signatures measured in parallel is being explored.

Figure 2 .
Figure 2. Simulated elemental () response functions for H, C, O, Si and Ca.

Figure 3 .
Figure 3. Mass fractions of H, C, O, Si and Ca unfolded from the transmitted neutron signatures using MAXED for composite media (solid).In each case the known ratio (hashed) is included for comparison.

Figure 4 .
Figure 4. Neutron fluence spectra unfolded with MAXED.The ISO recommended spectrum for AmBe is included for comparison.

Figure 5 .
Figure 5. Measured signature Σ  for neutrons transmitted through 6.0 cm SiO2, shown with the refolded signature obtained with an energy range of 1.0 -8.0.MeV.