Am-241 thermal neutron capture cross section and neutron capture resonance integral from reactor activation and oscillation measurements

. In order to validate the neutron capture cross section in 241 Am, comparisons of calculations with activation experiments at the TRIGA reactor at JSI and pile oscillation measurements at the MINERVE reactor at CEA Cadarache were done. The results suggest that the evaluation in the new test library JEFF-4T1 on average provide an improvement of the C / E results compared to the existing nuclear data library releases. However, due to some discrepancies, additional experiments at JSI as well as further development of the computational model of the JSI TRIGA reactor are planned to improve the accuracy and reliability of the neutron activation analysis and the nuclear data inferred from them.


Introduction
In spent nuclear fuel, 241 Am is an important contributor to the decay heat and radiotoxicity for cooling times between several decades and a few centuries. Inherent γray emission from 241 Am aggravate time-of-flight (TOF) measurements of 241 Am capture cross section. Alternatively, 241 Am thermal neutron capture cross section and resonance integral may be estimated by analysis of activation or pile oscillation measurements. Potentially, the former may be more accurate than the one obtained from TOF measurements. It may thus serve for normalisation of the energy dependent capture yields measured by TOF.
Studies within the framework of the Working Party on International Nuclear Data Evaluation Co-operation (WPEC) Subgroup-41 (SG-41) of the Nuclear Energy Agency (NEA) have shown that there are significant biases in the derivation of the thermal 241 Am capture cross section from activation measurements with cadmium transmission filters using the conventional Westcott method. It was shown that this problem can be overcome either by a higher order correction to the Westcott method [1] or by Monte Carlo calculated correction factors [2].
Neutron activation analysis of 241 Am is comparatively complex. First, 241 Am cross section contains two resonances below and around the cadmium transmission filter cut-off energy (∼ 0.55 eV). Second, the activation product is produced in both ground and metastable state, and the latter has a much longer half-life (141 y for 242m Am vs. ∼ 16 h for 242g Am). And finally, the decay scheme of activation products is relatively complicated. γ-ray spectrometry of activation products is difficult due to low γ-ray energies and increased γ-ray background from 241 Am fission products and 241 Am itself. Therefore, the only realistic option is α-particle spectrometry. α-particles originating from 242 Cm, which is a decay product of 242g,m Am, are measured relative to α-particles originating from 241 Am. Due to the much longer half-life of 242m Am and a relatively small branching fraction for its production by neutron capture (∼ 0.09), its contribution to the 242 Cm activity is negligible compared to the contribution from 242g Am for a few years after irradiation.
In contrast, pile oscillations are performed by introducing a sample containing 241 Am into the reactor during operation, and the resulting decrease in system reactivity is assumed proportional to the neutron capture rate in 241 Am. Compared to activation measurements, the required amount of 241 Am in the samples is several orders of magnitude higher.

Neutron activation experiments at the JSI TRIGA reactor
Two 241 Am samples, one with and one without cadmium cover, were separately irradiated in each of the following irradiation channels of the TRIGA reactor at Jožef Stefan Institute (JSI) [3]: Central Channel (CC), irradiation channels F-19 and IC40 (figure 1). These three irradiation channels were chosen due to very different thermalto-epithermal spectral ratios [4]. The neutron fluence was monitored by reactions 59 Co(n,γ), 197 Au(n,γ) and 58 Ni(n,p).
For the irradiations in CC, the count rates as well as other data regarding neutron irradiation and activation   [5]. For the irradiations in channels F-19 and IC40, the count rates will be published in another document. The first step is to determine the specific reaction rates starting from the measured detector count rates.
The reaction rates were deduced from α-particle spectrometry ( 241 Am) and γ-ray spectrometry (other materials). The 241 Am(n,γ) 242g Am reaction rates are summarised in table 1, whereas the other reaction rates are presented in table 2. A full derivation procedure including uncertainty propagation is described in detail in Ref. [5].

C/E comparison of reaction rate ratios
In parallel, the reaction rates were calculated using a 3D full-core computational model using Serpent 2.1.31 [6]. A relative C/E comparison is given in table 3.
The reaction rate ratios of Au, Co and Am samples irradiated with/without Cd filter provide a measure of quality of spectral ratios. The reaction rate ratios of the threshold reaction 58 Ni(n,p) in irradiated Ni samples provide a measure of total neutron fluence since only minor disturbances through the Cd filter are expected.
In CC, the spectral ratios are well described, whereas the total fluence was lower for the irradiation without Cd, which might be a consequence of local depression of the fission rates in the surrounding fuel due to thermal neutron absorption in Cd. This requires a further investigation.
In F-19, the total fluence was much higher for the irradiation with Cd filter, which is unexpected. The most likely reason is the uncertainty of the sample placement within the irradiation channel in combination with large radial gradients of the neutron field within the channel. Most likely, this experiment needs to be repeated ensuring a more precise sample placement.
In IC40, the total fluence between the two irradiations is in agreement. However, there is a known problem of the spectral ratios in the computational models of the TRIGA reactor of 10-20% [7].
The reaction rate ratios of Am/Au provide a measure of quality of 241 Am(n,γ) cross sections in different libraries. This is reliable only if both the total fluence and the spectral ratio are in agreeement.
For the thermal part of the neutron spectrum, which is important for interpretation of the measurements without Cd filter, this condition is met only for CC. It can be observed that using nuclear data (ND) libraries ENDF/B-VII.0 and ENDF/B-VII.1, in which the thermal neutron capture cross section values for 241 Am are 619 b and 684 b, respectively, crudely underestimates the Am/Au reaction rate ratio. On the other hand, the value of σ(E th ) = 748 b from JEFF-3.3 ND library leads to a slight overestimation of the Am/Au reaction rate ratio. The recent test ND library JEFF-4T1 (with σ(E th ) = 717 b) is in best agreement with the experiment.
For the resonance and fast part of the energy spectrum, the full-core computational model provides a better description of the reality, which is understandable since compared to thermal neutrons the fast neutrons undergo less collisions starting from the fission source. The Am/Au reaction rate ratios under Cd filter are consistent for all irradiation channels. For ENDF/B-VII.0 and ENDF/B-VII.1, the calculated Am/Au reaction rate ratios are significantly below the measured ones, which implies significantly underestimated values for the resonance integrals within these libraries. On the other hand, the C/E values (i.e. the ratios between the calculated and measured values) are consistent in all irradiation channels for JEFF-3.3.

Pile oscillation measurements at the MINERVE reactor at CEA Cadarache
The underestimation of the neutron capture cross section of 241 Am was also the subject of intense experimental activities at the CEA (Commissariat à l'énergie atomique et aux énergies alternatives) of Cadarache. Several    (12)  experiments were performed in the zero power reactor MINERVE, which was designed to measure the reactivity worth of small samples by using the pile oscillation technique. Integral trends, expressed in term of C/E-1 results, were obtained in the framework of the OSMOSE and AM-STRAMGRAM programs.
The OSMOSE program consisted to measure the reactivity worth of 10 cm long cylindrical samples composed of a stack of UO 2 pellets containing actinides in a doublesealed zircaloy container. The diameter of the samples was close to 1 cm. Two samples containing different amount of 241 Am were measured in UOX and MOX core configurations. The analyses were performed with the deterministic code APOLLO2 and the Monte Carlo code TRIPOLI4. The results are reported in Refs. [8][9][10].
The AMSTRAMGRAM program was specifically designed to complement the OSMOSE results by measur-ing the reactivity worth of 241 Am samples in a thermalised neutron spectrum. Oscillation measurements were performed with one of the OSMOSE sample, together with samples composed of a stack of Al 2 O 3 pellets containing AmO 2 provided by JRC (Joint Research Centre) Geel. Each pellet has a diameter of 1.2 cm and a height of about 2 mm placed in a Al container. Results obtained with the Monte Carlo code TRIPOLI are reported in Ref. [11,12].

Discussion
Solving the underestimation of the 241 Am(n,γ) cross section observed with the past neutron libraries was a long and tedious work. Owing to the expected correction as large as +20%, the purpose of the evaluation activities was to identify biases in the microscopic and integral data used in the evaluation procedure. NEA WPEC SG-41 succeeded to propose a thermal capture cross section lying between 702 b and 716 b. The results delivered by the analysis of the JSI data favour the lower value, close to 700 b. For the capture resonance integral, the combination of the MINERVE and JSI trends suggests a slight decrease of the 241 Am(n,γ) reaction above the Cadmium cut-of energy. The expected value could range from 1750 b to 1800 b. An Integral Data Assimilation of all the available data is needed in order to provide a better estimation by taking into account common uncertainty components for different experiments.

Conclusions
The experimental and evaluation works on the 241 Am capture cross section are well documented in the literature. However, the reported results highlight large discrepancies. The collaboration between JSI and CEA Cadarache aims to tackle this issue thanks to the experimental capabilities at the TRIGA reactor of JSI. Results obtained in the Central Channel confirm the integral trends obtained in the MINERVE reactor located at CEA Cadarache and suggest that the new 241 Am cross section evaluation available in the test library JEFF-4T1 (σ(E th )=717 b, I 0 =1826 b) will improve all the C/E results.
A closer inspection of the results still indicates some differences, not fully solved with the JEFF-4T1 test library. In the Central Channel, measurements with and without Cd still indicate a remaining overestimation of the C/E by 3.2% and 2.2%, respectively. These trends are in the upper part of the experimental uncertainties. They provide a good opportunity to explore the experimental accuracy that can be achieved at the TRIGA reactor for activation measurements. Therefore, additional experiments are planned at JSI with an improved experimental set-up that will be used in the F-19 and IC40 channel.