The scintillation light was transmitted to photomultipliers at the ends of the counters by wavelengths shifting lightguides and a photon-counting system was used to discriminate between neutrons and other events .sx A stack of eight of these counters , around a central cavity , formed the multiplicity detector .sx While the neutron counters were designed for a very specific purpose , it is thought likely that high efficiency low background counters of this type might find applications in other areas of nuclear instrumentation .sx In particular , arrays of neutron counters optimized for neutron multiplicity measurement have proved useful in the non-destructive assay of fissile materials ( Prosdocimi 1979) .sx 2 .sx Design of the neutron multiplicity detector .sx An experimental choice had to be made between designing a detector that could detect a number of fast neutrons simultaneously or designing a detector that first thermalized the neutrons in a moderator and then detected the thermal neutrons over a period determined by their diffusion time .sx Since the cross sections for typical fast neutron induced reactions are orders of magnitude lower than those of the corresponding thermal reactions , fast neutron detectors have much lower detection efficiencies .sx While they are more efficient , detectors based on thermalization can give no information about the initial energy of the neutrons .sx For the purposes of the current experiment , this was not considered a disadvantage .sx The previously mentioned requirement to achieve high efficiency and low background had also to be balanced against financial considerations .sx Undoubtedly an array of high pressure 3 He proportional counters in a large volume of heavy water would prove the most favourable arrangement , but the cost of such an array is prohibitive .sx Using a hydrogenous moderator , rather than one based on deuterium , enables the system to be much smaller but the efficiency is then limited by the magnitude of the ( n,p ) capture cross section .sx One design of neutron detector uses the 2.3 MeV gamma ray from this reaction to register the neutron .sx Such an arrangement , in which a liquid scintillator also acts as the moderator , is simple to implement but the high background rate prevents its use in many applications , including the present one .sx A similar , though less severe , restriction also applies to organic scintillators loaded with a small proportion of cadmium or gadolinium .sx Tanks of gadolinium loaded liquid scintillator formed the detectors used by Cheifetz et al ( 1972 ) and Becker et al ( 1979) .sx While extremely high neutron detection efficiencies were achieved ( 65% and 75% respectively ) , both these detectors were sensitive to environmental background gamma radiation .sx To reduce this sensitivity , Becker et al ( 1979 ) required the detection of fission fragments in coincidence with the neutrons before an event was registered .sx Consequently their detector could only be used with extremely thin samples from which the fission fragments could escape .sx Because of the above considerations , other neutron multiplicity detectors devised for superheavy element searches have used an array of 3 He proportional counters in a matrix of hydrogenous moderator ( Macklin et al 1972 , Ter-Akop'yan et al 1981 ) and this approach has also been adopted for fissile material assay ( Prosdocimi 1979) .sx The neutron multiplicity detector described here differs from all the previous detectors in that the detection of the neutrons is achieved by the scintillation of a lithium fluoride and zinc sulphide mixture and that the hydrogenous moderator is a structural component of the individual neutron counters .sx The neutron capture reaction in 6 Li produces both a triton and an alpha particle , with an energy release of 4.79 MeV .sx Silver-activated zinc sulphide is well known as the most efficient of all scintillators and consequently intimate mixtures of the two materials coupled to photomultipliers are well established as cheap and efficient slow neutron detectors .sx The main limitation of such mixtures is that , as with other zinc sulphide scintillators , the material is opaque to its own light and consequently can only be used in thin layers .sx This renders it difficult to construct large volume detectors .sx It was shown by Barton and Caines ( 1970 ) that this limitation could be overcome by the use of wavelength shifting lightguides .sx This technique is used again in the present detectors with a slightly different geometrical arrangement .sx 3 .sx Construction of the neutron counters .sx The neutron counters had an overall sandwich construction consisting of thin layers of zinc sulphide scintillator attached to the slabs of polypropylene moderator that provided the major part of the moderating material .sx These slabs were interleaved with the wavelength shifting lightguides , leaving a narrow air gap between the neutron scintillator layers and the lightguides .sx The light emitted by the scintillator traversed the lightguides causing fluorescence and some of the re-emitted light was then carried along the guides .sx Both ends of the lightguides were viewed by photomultipliers , there being a second air gap between the lightguides and the faceplate of each photomultiplier .sx The whole counter was encased in a light-tight black Perspex tube of square cross section .sx The photomultipliers were held in head units constructed from diecast boxes bolted to the ends of the Perspex casing .sx A longitudinal section of one of the counters is shown in figure 1 , while figure 2 shows a cross section .sx The zinc sulphide chosen had a larger than usual proportion of nickel added to reduce the fraction of light in the tail of the scintillation pulse ( Barton and Ranby 1977) .sx It was prepared by Thorn Research and Engineering Laboratories as HS338A .sx The lithium fluoride was material that had been enriched to 96% in 6 Li by Harwell Stable Isotopes Group .sx The grains of both powders had a most probable size of about 5 mu m but the range of sizes was not closely controlled and included grains from less than 1 to 10 mu m. The two powders were thoroughly mixed together , then added to Dow Corning Sylgard 184 .sx This material is a thermosetting silicone resin which is rubbery and optically transparent when set .sx The mass proportions were ZnS :sx LiF:resin:hardener = 4 :sx 2:2.7:0.3. The resulting mixture was used to coat sheets of highly reflective aluminized Melinex .sx Considerable effort was expended in finding a suitable technique to manufacture large areas of the scintillator with a uniform thickness .sx The technique finally adopted was to spread the mixture which , before setting , had the consistency of thick paint .sx A sheet of plate glass was mounted horizontally on the bed of a milling machine to provide a flat working surface and the sheet of Melinex to be coated was placed on top of this .sx The scintillator mixture was loaded into a slotted hopper which was mounted on the transverse of the machine and this hopper was pulled slowly across the Melinex by moving the bed below , spreading the mixture as it went .sx The intention was to provide layers of thickness 100 unch 10 mu m but in practice the variation , as measured by a gamma-ray absorption technique , was slightly more than this .sx The only other disadvantage of this technique was that the wastage of material was about 25% .sx figures&captions .sx The incorporation of such a high proportion of foreign chemical material in the silicone resin increased its setting time considerably and the surface of the layers remained permanently sticky .sx When the first prototype counters were tested in an underground laboratory , their background counting rates were observed to increase slowly from their initial values .sx Tests showed that this effect was due to radon decay products , either as atoms or attached to dust particles , settling on the surfaces of the scintillator layers .sx The problem was solved by applying a second coating of pure silicone resin over the scintillator .sx This second layer , which set perfectly , was of sufficient thickness to prevent the alpha particles from any subsequently deposited atoms from reaching the scintillator but did not interfere with the light leaving the scintillator .sx The scintillator layers , on their Melinex backing sheets , were attached to the polypropylene blocks by double sided adhesive tape .sx The wavelength shifting lightguides were fabricated from R o hm and Haas Plexiglas GS2025 , a polymethyl - methacrylate glass containing the fluorescent compound BBQ .sx All the surfaces were polished and the guides were supported away from the sides of the Perspex casing of the counter using conically tipped screws .sx The array of guides was viewed from each end by a 130 mm photomultiplier with aluminium foil providing a simple lightguide .sx The photomultipliers were EMI type 9791KA tubes .sx They were individually selected to have a low afterpulsing rate , specified as less than 400 ion-induced pulses , of magnitude equivalent to more than 2.5 single photoelectrons , for every 10 6 photoelectrons .sx figure&caption .sx The external dimensions of a complete counter , excluding the photomultiplier head-unit were 900 mm x 144 mm x 144 mm .sx A stack of eight of these counters could be arranged as shown in figure 3 .sx The head-units have been shown on one counter only and omitted on the others for clarity .sx This arrangement of counters provided a central cavity for samples with the same dimensions as a single counter ( volume 18.61 ) , which could adequately contain several tens of kilograms of sample material .sx The central detector in the top row , immediately above the cavity , was kept in position by two strips of aluminium sheet , each a few centimetres wide , below it .sx These strips were placed close to the two ends of the stack across the two detectors which comprised the second row and before the top row was added .sx 4 .sx Output pulse processing and event recording .sx The pulse-height distribution of neutron-induced scintillation in lithium loaded zinc sulphide scintillator is such that simple pulse-height discrimination between neutrons and noise signals is not feasible .sx This is a consequence of the particular time structure of the scintillation in zinc sulphide and of the wide range of pulse heights resulting from both the heterogenic nature of the photon source and variations in light collection and conversion .sx Many types of pulse shape discrimination circuit have been reported but none seemed suitable for the rather stringent requirements of the present experiment .sx In addition to the desire to maximize efficiency , there were also overriding requirements that photomultiplier and electronic noise should not contribute to the background counting rate and that the system should discriminate against signals due to the passage of one or more lightly ionizing particles .sx The method finally adopted was a digital pulse discrimination system with two distinct stages .sx The first of these was a photon counting discriminator , based on earlier work by Caines ( 1972 ) and Davidson ( 1977) .sx A block diagram of it is shown in figure 4 ; exact details of the circuits can be found in McMillan ( 1990) .sx Firstly , the output of each photomultiplier was converted to a sequence of uniform pulses by a simple monostable circuit based on a fast voltage-comparator and having a pulse resolution of 40 ns .sx The input level of this monostable was set so that it was triggered by every photoelectron .sx Noise and most minimum ionizing events were rejected by requiring that there was an initiating prompt coincidence between signals from the photomultipliers at both ends of the counter and that there was then a minimum of two other pulses from each end within the next 20 mu s. After this period there was a dead-time of a further 20 mu s to ensure that single neutrons could not masquerade as doubles .sx If the above criteria were met , the circuit produced a 'neutron output' signal .sx Additionally , a pulse-train output was generated by combining the outputs from the discriminators associated with the two photomultipliers in an OR gate .sx The exact timing and selection criteria chosen were inevitably a compromise between rejecting some neutron events and accepting some non - neutron events .sx The first stage , as described and with the values indicated , provided neutron discrimination sufficient for most purposes .sx The efficiency and neutron lifetime measurements described later were performed using only this first stage .sx However , the fact that it failed to reject certain minimum ionizing events , particularly small electromagnetic showers , meant that for low background multiplicity work an additional level of discrimination was necessary .sx figure&caption .sx This second stage of pulse processing was not so much an electronic discrimination system as a method of recording the signals from the counters in sufficient detail that minimum ionizing events could subsequently be recognised by eye in the data set and rejected accordingly .sx