The effects of a revised operational dose quantity on the response characteristics of neutron survey instruments

According to the proposed definition [6, 7], the fluence to dose conversion coefficient at a given energy, E_max/Φ (or equivalently H*/Φ), will be set equal to the maximum of the corresponding fluence to effective dose (E_G/Φ) conversion coefficients at that energy taken over all geometries, G, considered by the International Commission on Radiological Protection (ICRP) [8], i.e. G denotes anterior-posterior (AP), posterior-anterior (PA), left-lateral (LLAT), right-lateral (RLAT), isotropic (ISO), semi-isotropic superior (SS-ISO), semi-isotropic inferior (SI-ISO), and rotational (ROT). A comparison of E_max/Φ and H*(10)/Φ as a function of neutron energy is shown in figure 1. Also shown in figure 1 is the quantity E_AP/Φ, which in comparison to H*(10)/Φ demonstrates the motivation for the proposed new quantity.

Zoom In Zoom Out Reset image size

For most of the energy range over which neutron survey instruments are often employed, i.e. from thermal to ∼20 MeV, H*(10) over-estimates the protection quantity; the magnitude of this discrepancy can vary from a ∼few percent between 1 and 10 keV, to a factor of ∼few at thermal energies. Although designing neutron survey instruments to respond in terms of H*(10) therefore provides an inherent degree of general conservatism in their routine use, in some workplace fields it can prove over-cautious. Moreover, for the energy range from ∼2 to ∼10 MeV, which is also close to the mean energies of the ²⁴¹Am-Be or ²⁵²Cf sources typically used to calibrate them, and above ∼40 MeV, assessment of H*(10) significantly under-estimates the protection quantity, and hence the potential risk to the individual. Conversely, E_max/Φ is always conservative relative to E_G/Φ, by definition, for all G. Further, unlike H*(10) for which the conservatism is somewhat 'artificial', the conservatism in E_max is arguably more grounded: E_max faithfully represents the largest possible value of effective dose in a given field for an individual orientated within it. On balance, therefore, for the overall meV to GeV energy range the proposed quantity E_max represents a more reliable assessment of potential risk to individuals than H*(10), with the advantages particularly significant in workplace fields featuring high energy neutrons.

For the purposes of the present exercise, four contemporary neutron survey instruments have been considered: the Berthold LB6411 [9], Studsvik 2202D [10], HSREM [11] and GNU [12]. There are many neutron survey instruments that could have been selected, but these were favoured because the GNU was recently designed at Public Health England (PHE), the LB6411 is widely used throughout Europe and is of relatively recent design, the Studsvik 2202D is representative of the Andersson-Braun family of instruments [13], and the HSREM is the newest variation of the Leake family of instruments [14, 15] based on relatively light spheres. The current manifestation of the Studsvik 2202D is the KWD Nuclear Instruments AB 2222A. The GNU, or 'Guided Neutron Unit', has recently been developed and is now commercially available: it has a flatter energy-dependence of H*(10) response than the other devices, is highly sensitive, and has a considerably better response to high-energy neutrons than instruments of comparable mass (∼8 kg, excluding electronics).

The dose response characteristics of these four devices have been calculated previously at selected neutron energies across a range spanning from meV to MeV, using H*(10)-per-fluence data applied to counts-per-fluence results obtained from Monte Carlo modelling; for the GNU, this energy range has been extended up to 1 TeV. The response characteristics of the instruments at those energies are now recalculated using E_max/Φ data in place of H*(10)/Φ.

In addition to their energy-dependent responses to monoenergetic sources, the performances of the four instruments in workplace fields have also been calculated and contrasted under the assumptions of the proposed change. This was achieved by convolving the respective fluence response characteristics of the devices with the 19 realistic workplace neutron fluence-energy distributions that were determined during the EVIDOS project [16], using a specially designed algorithm that 'folded' the relevant data pointwise across a suitably fine energy grid. Each result was then divided by the mean H*(10)/Φ or E_max/Φ (as appropriate) for that field, similarly derived by folding the distribution with the monoenergetic conversion coefficient data.

The H*(10) or E_max response of each instrument to²⁴¹Am-Be and ²⁵²Cf calibration sources were also calculated by means of the same folding method, using reference fluence-energy distribution data [17].

The absolute H*(10) and E_max responses of the LB6411, Studsvik 2202D, HSREM and GNU are shown in figures 2(a) and (b) respectively, with data for the GNU presented both for isotropic exposures and for a plane-parallel exposure orientated opposite to the stem of its central, ³He-filled, LND detector [12]. The same data are presented in figures 3(a) and (b), but shown relative to each instrument's calculated H*(10) or E_max response, as appropriate, to ²⁴¹Am-Be. The absolute H*(10) and E_max responses of the four devices to ²⁴¹Am-Be and ²⁵²Cf sources are given in table 1.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

It is seen in figures 2(a) and 3(a) that the GNU has the best H*(10) response for thermal neutrons and a relatively small over-response around 10 keV. However, this advantage would be reduced by a change to E_max: the responses of all of the instruments are flatter below ∼10 keV (figures 2(b) and 3(b)), but the change in the conversion coefficients for thermal neutrons gives the GNU a poorer response at that energy (figure 3(b)).

For the LB6411 and Studsvik 2202D, amending the dose quantity would lead to a slightly flatter energy-dependence of response, although the Studsvik is still less sensitive than the LB6411, which is in turn considerably less sensitive than the GNU. The response of the highly sensitive HSREM is also arguably improved by a change from H*(10) to E_max, though it is still more energy-dependent than the other designs, which is the penalty of being the lightest detector. For all four instruments, the differences are most manifest at low energies, though the 'dip' in H*(10) response at around 100 keV is also flattened by the change to E_max; the change neither improves nor exacerbates the significant over-response peaked at ∼5 keV that is typical in many designs of neutron survey instrument. Of course, these energy-dependent trends are as anticipated from figure 1, and result in part from the limitations in the use of H*(10) as a surrogate for the protection quantity, rather than from deficiencies in the designs of the instruments themselves.

In all cases, the relative responses of the instruments could potentially be improved by adopting a different calibration. This might be achieved by scaling the ²⁴¹Am-Be response by a constant factor, chosen so that it provides a more optimal response characteristic across the energy range. An alternative approach might be to recalibrate to a different energy or source, though routine checking of the instrument would likely still need to use a readily available radionuclide such as ²⁴¹Am-Be or ²⁵²Cf. Examples of this possibility are shown in figure 3(b), where the relative performance of the GNU is seen to be improved by renormalizing to its responses at either 144 or 565 keV, which were determined by interpolating the monoenergetic data (figure 2(b)). These calibration energies can readily be provided by most Primary Standards Laboratories for neutron exposures, for instance the UK's National Physical Laboratory (NPL) [18]. Indeed, a 'hybrid' approach might also be possible, such as routinely calibrating to ²⁴¹Am-Be, say, but using its ratio with the 565 keV response to inform the choice of optimum scaling factor. Ultimately, however, these options are just compromises: in the current context, figure 3(b) illustrates that although recalibration may improve relative responses overall, it could not improve the energy dependence of an instrument's response and so cannot alone solve all of the limitations that would be introduced for the current designs of instruments upon adopting the new dose quantity.

The performances of neutron survey instruments are typically judged against criteria stipulated in international standards [19, 20]. The IEC standard recommends that the relative responses of instruments should lie between 0.2 and 8.0 in the energy range from thermal to 50 keV, between 0.5 and 2.0 from 50 keV to 10 MeV, and between 0.2 and 2.0 from 10 MeV to 20 MeV. These limits are shown on figure 3(a), where it is apparent that the instruments had acceptable response in terms of these criteria (though the HSREM does exceed them slightly at the keV scale), but that they no longer always satisfy the requirements when responding in terms of E_max (figure 3(b)); of course, these limits would need to be shifted proportionally if the alternative calibrations to 144 or 565 keV shown in figure 3(b) were adopted. However, the IEC criteria are given in terms of the H*(10) response, so it may be argued that they would not be an appropriate test for instruments calibrated to E_max. This leads naturally to the suggestion that changing the dose quantity from H*(10) to E_max would require corresponding amendments to the standards. For example, it may no longer be appropriate to permit a factor of 5 under-response for thermal neutrons, an important energy in many workplaces. This was allowed previously because, in terms of H*(10), there needed to be a balance between the under-response for thermals and the over-response around 10 keV. At the other extreme of the energy range, the high energy response of the GNU is seen to be significantly poorer for E_max than for H*(10), which will be hard to remedy without the addition of considerably more lead and hence more mass. The other instruments, which do not have response data available for high energies, would have poorer responses than the GNU, so they are unlikely to have acceptable performances above 20 MeV. This could cause problems for monitoring in workplace fields with high energy components, such as around accelerator or fusion facilities.

When deciding upon the optimum calibration for each instrument, if they are to be judged against the same IEC 61005:2014 [19] criteria, there are clearly a number of problematic energy ranges. Thermal neutrons will not present the difficulty that they did previously, but there will still be a potential for exceeding the IEC criteria at around 10 keV. Moreover, excessive response around 10 keV will make it hard for the design to meet the criteria at 50 keV. It is already hard for instruments to achieve a response >0.5 at 10 MeV, but this will be more difficult with E_max if they have no high-Z layer. Adoption of a different calibration to solve the response issue around 50 keV will make the 10 MeV issue more severe. Ultimately, it will be for the manufacturers of the instruments to consider carefully the calibration implications for their devices, with some compromise inevitably required. The examples given in this section are provided just for the purposes of discussion, and should not be misinterpreted as how each instrument would have to be calibrated if the new operational quantities were adopted.

The H*(10)/Φ conversion coefficients for the 19 EVIDOS fields [16] were all larger than their E_max/Φ counterparts: an average ratio of ∼1.6× was found, with the largest nearly a factor of 2 (table 2). This is because the workplace fields are dominated by components with energies less than a few MeV, for which H*(10) provides excessively conservative estimates of E (figure 1). Regardless, the observation demonstrates the point that changing the dose quantity will effectively change the perceived characteristics of radiation fields, and not just the responses of instruments used within them.

The H*(10) and E_max responses of the instruments to the 19 EVIDOS workplace fields are shown in table 2, given relative to their respective responses to ²⁴¹Am-Be. The mean of each set of H*(10) or E_max responses for the fields is also given for each instrument, along with the standard deviations of those 19 results around their average. The performances of the instruments can be seen to differ to greater or lesser extents, depending on the energy characteristics of both the fields [16] and the instrument responses (figures 2 and 3). However, one obvious trend is that the relative E_max responses of the four instruments in the workplace fields are all greater than their relative H*(10) responses, which is expected from the comparison of the H*(10)/Φ and E_max/Φ conversion coefficients. A second obvious trend is that the 19 H*(10) response results for a given instrument are generally distributed closer to their mean than those for E_max, with wider dispersals typically evidenced in the latter datasets. This weaker consistency for E_max relative to H*(10) is caused not just by differing response characteristics for the instruments, but also by differing characterizations of the fields themselves due to the amended dose quantity.

For a given instrument, the mean response value (table 1) could be reinterpreted as the factor by which the response should be increased relative to that for ²⁴¹Am-Be to achieve optimum performance in the workplace. However, the selection of EVIDOS fields used does not include any very hard distributions, so it does not seem appropriate to make assumptions about how the manufacturers might individually choose to deal with this issue.

The LB6411 over-responds the least to E_max, but always under-responds to H*(10) for this selection of fields, in part because of its low response to thermal neutrons. A second cause of the under-response is the choice of normalization: for the LB6411 it is typically preferable to calibrate to ²⁵²Cf rather than ²⁴¹Am-Be. However, although it can be shown from the data in tables 1 and 2 that this would improve the under-response to H*(10), with the mean relative response increasing from 0.83 for ²⁴¹Am-Be to 0.86 for ²⁵²Cf, for example, there is still a significant over-response to E_max, although the mean relative response would decrease from 1.43 for ²⁴¹Am-Be to 1.24 for ²⁵²Cf.

The GNU and Studsvik are comparable in performance, over-responding to E_max by average factors of ∼2.0× and ∼1.8× respectively. The HSREM significantly over-responds to E_max, but it also over-responds to H*(10) in 12 out of the 13 fields.

In all cases, the over-responses shown in table 2 could potentially be reduced by routine recalibration to a source other than ²⁴¹Am-Be, or by rescaling the ²⁴¹Am-Be calibration by a constant factor. Similarly to before (figure 3(b)), examples of such alternative calibration sources might be 144 keV or 565 keV, both of which would roughly halve the results. To illustrate this possibility, the E_max responses of the GNU to workplace exposures have been renormalized to its responses at both 144 keV and 565 keV. The recalibrated responses are shown in figure 4, alongside the E_max and H*(10) results for the usual ²⁴¹Am-Be normalization (table 2); each datapoint is plotted at the mean H*(10)/Φ or E_max/Φ (as appropriate) for that field. The figure indicates relative pros and cons between using either 144 keV or 565 keV calibrations for E_max, but both clearly produce datasets that are generally much closer to unity than that for ²⁴¹Am-Be. Of course, as before (figure 3(b)), these recalibrations are chosen to demonstrate general trends and are not necessarily considered optimal.

Zoom In Zoom Out Reset image size