Time- and energy-resolved effects in the boron-10 based Multi-Grid and helium-3 based thermal neutron detectors

The boron-10 based Multi-Grid detector is being developed as an alternative to helium-3 based neutron detectors. At the European Spallation Source, the detector will be used for time-of-flight neutron spectroscopy at cold to thermal neutron energies. The objective of this work is to investigate fine time- and energy-resolved effects of the Multi-Grid detector, down to a few $\mu$eV, while comparing it to the performance of a typical helium-3 tube. Furthermore, it is to characterize differences between the detector technologies in terms of internal scattering, as well as the time reconstruction of ~ $\mu$s short neutron pulses. The data were taken at the Helmholtz Zentrum Berlin, where the Multi-Grid detector and a helium-3 tube were installed at the ESS test beamline, V20. Using a Fermi-chopper, the neutron beam of the reactor was chopped into a few tens of $\mu$s wide pulses before reaching the detector, located a few tens of cm downstream. The data of the measurements show an agreement between the derived and calculated neutron detection efficiency curve. The data also provide fine details on the effect of internal scattering, and how it can be reduced. For the first time, the chopper resolution was comparable to the timing resolution of the Multi-Grid detector. This allowed a detailed study of time- and energy resolved effects, as well as a comparison with a typical helium-3 tube.


Introduction
The European Spallation Source (ESS), is currently under construction in Lund, Sweden [1,2,3]. Due to the sparsity and increased cost of helium-3 gas during the past two decades [4], as well as the performance limitation due to the expected high flux from the source, alternatives to the traditional helium-3 based neutron detectors are being developed. One such alternative is the boron-10 based Multi-Grid detector [5,6,7,8,9,10,11], invented at the Institut LaueLangevin (ILL) [12], and jointly developed with ESS thereafter. The Multi-Grid detector is a large area cold to epithermal neutron detector, designed for time-of-flight neutron spectroscopy [13] at the upcoming CSPEC [14] (cold neutron spectroscopy) and T-REX [15] (thermal neutron spectroscopy) instruments at ESS.
Neutron spectroscopy is a demanding technique in terms of neutron detector performance, requiring a high detection efficiency and low noise levels. It also needs a broad detector coverage, tens of square meters, and hence a low cost per unit active detector area is important. The spectroscopy information is primarily extracted from inelastic neutron scattering of a sample, measuring small changes in energy between incident and scattered neutrons. This probes sample properties such as molecular vibrations, quantum excitations and motions of atoms. As the measured energy changes are in the order of meV, the method requires a high time resolution and good knowledge of interaction location to correctly derive the energy of the scattered neutrons [16,17,18].
For neutron detection, the Multi-Grid detector employs multiple 10 B 4 C thin films coated on aluminum substrates [19,20,21]. The 10 B 4 C deposition is carried out by the ESS Detector Coatings Workshop in Linköping, where the deposition is done with an industrial deposition system, using physical vapor deposition DC magnetron sputtering. The coated substrates are stacked perpendicularly to the incident neutrons, and placed in a 3D-position sensitive multi-wire proportional chamber (MWPC), see Section 2.4 for details. Incident neutrons are absorbed in the coating, whereupon one of the conversion products is ejected into the gas volume. In the gas, charge is released and collected, and each neutron event is assigned a time-of-flight (tof ), as well as a hit position. This information is then used to calculate the neutron flight distance (d).
The neutron time-of-flight and flight distance are relative to a neutron chopper. A neutron chopper is a device containing an efficient neutron absorber, such as gadolinium or boron, and is placed blocking the incident neutron beam. The chopper rotates at a high frequency, and contains an opening for neutrons to pass periodically. That is, the incident neutron beam is "chopped" into pulses. Each time a neutron pulse is transmitted, a start time, commonly know as T 0 , is recorded. Then, when a neutron is detected in the detector, a stop time T is acquired. The time-of-flight is then calculated according to tof = T − T 0 , and the neutron flight distance d is the distance between the chopper and the detection location in the detector, i.e. the hit position.
Using the time-of-flight and flight distance, the incident neutron energy E n can be calculated on an event-by-event basis. This is done by calculating the neutron velocity v n , according to v n = d/tof , which gives the energy using E n ∝ v 2 n . However, due to the aluminum present inside the Multi-Grid detector, a small fraction of neutrons undergo internal scattering before being detected. Consequently, these scattered neutrons are assigned an incorrect flight distance, as the additional scattered travel path cannot be accounted for. This changes the energy line shape reconstruction, adding a small distribution of neutrons with incorrectly assigned energies [22,23,24]. A good understanding of this effect is important, as spectroscopy instruments depend on a well understood energy line shape. This is especially true for quasi-elastic scattering analysis [25], where subtle details in the line shape are studied at the edges of the energy peak.
For cold neutrons, the detector time and energy resolution are also critical. The full resolution of the instrument depends on the pulse broadening from the chopper system, sample scattering and the detector. Therefore, it is desired to keep the detector energy resolution finer than the resultant resolution of the remaining components. This is especially important for the slowest neutrons, as the resolution of the chopper system is highest at these energies. One of the main components of the detector resolution relates to the uncertainty of when and where the neutron conversion reaction occurs in the detector. In a helium-3 tube, this reaction can happen anywhere within the gas, while for the Multi-Grid detector, the reaction can only occur at discrete intervals, corresponding to the position of the conversion layers [23].
This work contains three separate investigations, each characterizing an important aspect of the Multi-Grid detector. First, the neutron detection efficiency of the Multi-Grid detector is derived and compared to the theoretical prediction. Then, the magnitude of the internal scattering, as well as the effect it has on the line shape reconstruction, is investigated. This is done by comparing two different Multi-Grid prototypes, one with internal topological shielding and one without, and examining the difference in energy line shape. Finally, the time-and energy resolution of Multi-Grid detector and the helium-3 tube are accessed and compared.

Instrumentation and experimental setup
The measurements were conducted at the Helmholtz-Zentrum Berlin (HZB), at the BER II research reactor [26]. At the facility, a series of measurements were done at the V20 beamline [27,28,29]. During the course of the measurements, three different detectors were used: one 10 bar helium-3 tube and two variations of the latest prototype of the Multi-Grid detector. The setup consisted of a beam-monitor, slits and a lightweight fast rotating Fermi-chopper [30]. The detectors were situated just after the Fermi-chopper, as illustrated in figure 2.
Three direct beam measurements were conducted, one for each of the three detectors, keeping the rest of the setup constant. In addition to this, two background measurements were done, one for each of the Multi-Grid prototypes, where the direct beam was blocked by the helium-3 tube.

The ESS test beamline V20
The V20 is a cold to thermal neutron beamline. It was commissioned as an ESS test beamline, with a chopper system designed to mimic the long pulses which will be obtained at ESS. This is done by a pair of double disc choppers: a source chopper and a wavelength band (WB) chopper. Together, these deliver pulses at 14 Hz, where the wavelengths in each pulse range from approximately 1 to 10Å. The beam line is also equipped with wavelength frame multiplication (WFM) choppers, which, together with a pair of frame overlap (FOL) choppers, cut the long pulse from the source chopper to smaller bunches. This option was, however, not used for the herein reported work, and the corresponding choppers were parked in open position. Instead, the choppers were used in the "Basic Single Pulse Mode", as illustrated in figure 1.  Figure 1: Illustration presenting the layout of the V20 instrument, showing the main components and their locations. At 21.7 m from the moderator, the ESS source chopper is located, which, together with the WB chopper, cuts the flux into wide pulses at 14 Hz. There are also WFM and FOL choppers, above presented in blue, followed by an optional polarizer. This allows for two different modes, "Basic Single Pulse Mode" and "WFM Mode". For these measurements, single pulse mode was used, which is illustrated above. The optical benches is where the experiment specific setup is positioned. Picture and in-picture captions adapted from [27].
The current setup was assembled at the "Optical Benches" position, at the last part of the beamline. This is outlined in figure 2, where all the components and their locations are presented. First, a slit confines the cross-section of the incident neutron beam to a few cm 2 . The incident flux is then recorded by a low helium-3 pressure beam monitor, with a ∼ 10 −5 detection yield [31]. After a further slit collimation, halving the beam width, a Fermi-chopper is used to cut and shape the long incident pulse to a few tens of µs short pulses. Finally, the beam is collimated to a 14 × 60 mm 2 rectangular beam before reaching the detectors. This is either the helium-3 tube or the Multi-Grid detector. During the direct beam measurements with the Multi-Grid detector, the helium-3 tube was removed from the beam path. Then, during the background measurements, it was re-installed such that it was blocking the direct beam from reaching the Multi-Grid detector. Note that the detectors are placed in sequence, due to the limitation of the setup, and that the Multi-Grid detector is situated approximately an additional 50 % further downstream of the Fermi-chopper than the helium-3 tube. The signal-to-background ratio at the instrument is approximately 10 3 , and is influenced by gamma radiation levels and straying neutrons in the vicinity of V20. 30

Lightweight fast rotating Fermi-chopper
For the measurements, an experimental 120 mm high, 27 mm in diameter, lightweight Fermi-chopper was used. The purpose was to produce a series of very short, ∼ 10 µs, and sharp pulses in time, to study the energy dependent scattering within the detector between pulses. The Fermi-chopper is synchronized to the 14 Hz source chopper frequency with a clock multiplier of 35, resulting in a 490 Hz rotation frequency. The chopper uses two rectangular chambers for neutron transmission, and a gadolinium neutron absorber on the walls. The resulting transmission of the Fermi-chopper is presented in figure 3. In figure 3a, the time-of-flight spectrum before the Fermi-chopper is shown, collected using the beam-monitor, while in figure 3b the time-of-flight spectrum after the Fermi-chopper is shown, collected using the helium-3 tube.
The chopper system produces a series of sharp pulses. However, as can be seen in figure 3c, there are additional features introduced. First, odd and even peaks are alternating in intensity. This is due to an asymmetry in the two rectangular Fermichambers, appearing every half rotation of the chopper blades. This is further discussed in [30]. Second, two distinct side peaks are identified, suppressed by 1-2 orders of magnitude. The same side peaks are present in data from the Multi-Grid detector. Thus, this feature is not a detector dependent effect. Additionally, there are 2-3 orders of magnitude suppressed peaks appearing midway between the main peaks, coinciding with every quarter rotation of the chopper blades. This was shown to correspond to a small misalignment of the incident neutron beam, allowing neutrons to pass the side of the Fermi package every quarter rotation. These peaks does not cause a problem for the current analysis, as they are easily distinguishable from the main peaks. Counts (c) Figure 3: Time-of-flight spectra before and after the Fermi-chopper. The time offset is arbitrary. In (a), a spectrum collected with the beam monitor is presented, showing the wide pulse from the source chopper. In (b), a spectrum from the helium-3 tube is shown, illustrating how the chopper splits the ∼ 60 ms long source pulse into a series of ∼ 10 µs short pulses. In (c), a magnification of a portion of (b) is shown. On the log scale, it is clear that there are side peaks which are 1-2 orders of magnitude smaller than the main peaks.

Helium-3 filled proportional counter
For the measurements, a helium-3 filled proportional counter from Reuter-Stokes [32] was used. The tube has a total gas pressure of 10 bars, split between 9.85 bar helium-3 and 0.15 bar quenching gas. The active diameter of the tube is 25.4 mm, with a length of 305 mm, and it is enclosed in 0.51 mm walls of stainless steel. The full geometry is presented in figure 4. During the measurements, the tube was wrapped in Mirrobor shielding [33], leaving a small window for incident neutrons.  The tube is a non-position sensitive detector, operated at 1350 V. Gamma events were rejected using a constant charge discriminator threshold on all events. The readout system consists of a multi-channel analyzer, the FAST ComTeC MCA4 [34], which returns time-of-flight and collected charge for each neutron event. In case of a pile-up detection during the charge integration process, the event is also assigned a "pile-up flag". The shaping time was set to 1 µs.

The Multi-Grid detector
The Multi-Grid detector has a modular design, see for example [5] for a rigorous description of the detector technology. It consists of a series of identical building blocks called grids, as shown in figure 5a. In the prototype used here, each grid contains 21 layers of 10 B 4 C-coated aluminum substrates, 0.5 mm thick, which are called blades. These are placed in a sequence and oriented orthogonal to the incident neutrons. The first blade is single-side coated and the remaining twenty double-side coated. These are the normal blades (blue in figure 5a). Connecting the normal blades are five support blades, called the radial blades (red in figure 5a). The naming convention refers to how the blades are oriented related to the incident neutrons. Together, these blades form a grid of cells. The normal blades have three different coating thicknesses, 1 µm, 1.25 µm and 2 µm, where the blades with the thinner coating are at the front and the ones with thicker coating at the back. If the radial blades are coated, they are double-side coated with a 1.25 µm coating thickness. At the back of the grid, a 5 mm thick Mirrobor sheet is placed. The purpose is to absorb the remainder of the incident neutrons, thus reducing back scattering.
The grids are stacked in three separate columns, see (1) in figure 5b for an example of a column, where each column contains 40 grids. Between the cells in the grids, anode wires are stretched, one wire per stack of cells, see (2) and (3) in figure 5b. These wires are then connected to a high voltage, while the cathode grids are put to ground potential. By placing the wires and grids in an Ar-CO 2 (80:20 volume ratio) filled vessel, a MWPC is established. The grids, as well as the wires, are electrically insulated from each other. This results in a position resolution defined by the cell size, where each cell is 22.5 × 22.5 × 10 mm 3 . This is defined as a voxel, and there are a total of 9600 voxels in the current prototype (40 height × 4·3 width × 20 depth). The corresponding active detector surface area is approximately 0.24 m 2 .
The induced charge released from the boron-10 neutron capture reaction is collected by the wires and grids. Each neutron event is then assigned a 3D-position based on coincidences in time between signals from wires and grids. Each event is also assigned a time-stamp, corresponding to the time-of-flight from the source chopper to the location of that event within the Multi-Grid detector. Note that each wire is adjacent to at least two 10 B 4 C-coated aluminum substrates, four if the radial blades are coated, and that incident neutrons can be converted in either one. The information of which coating the conversion took place is not recorded. For the measurements, two separate prototypes of the Multi-Grid detector were used. These prototypes were originally designed and constructed as demonstrators for measurements at the thermal spectrometer SEQUOIA [17] in August 2018, at the Spallation Neutron Source (SNS) [35]. The prototypes are therefore optimized for thermal neutron energies. The two prototypes are identical in all aspects, except for the grid specifications, which are summarized in table 1. The grids in Detector 1 is made with radiopure aluminum from Praxair [36], with less than 1 ppb (parts per billion) radioactive impurities such as Th and U, while the grids in Detector 2 is made with commercially available Al5754. By limiting the amount of radioactive impurities present in the aluminum, the intrinsic background can be suppressed [10]. However, for these measurements, the parameter of interest is the 10 B 4 C-coating on the radial blades, see figure 5a. This coating is only present in Detector 2. The purpose of coating the radial blades is to reduce the effect of internally scattered neutrons. By adding the extra coating, scattered neutrons can be absorbed more quickly. This reduces the average extra flight distance by scattered neutrons, which in turn improves the accuracy of the energy reconstruction. In addition to this, the overall neutron detection efficiency for incident divergent neutrons, i.e. incident neutrons traveling at a path crossing the radial blades, is increased. This is due to the additional converter material introduced. As a consequence, neutrons are statistically absorbed closer to the detector entrance, reducing the travel path through aluminum within the detector. This lowers the number of interaction opportunities in aluminum, reducing the overall amount of internally scattered neutrons. This further improves the accuracy of the energy reconstruction.
The read-out electronics is the mesytec VMMR-8/16 [37], which can handle 2048 channels simultaneously. In this setup, 360 channels are used, corresponding to the 80 · 3 = 240 wires and 40 · 3 = 120 grids. The data is transported using optical fibres. Each recorded event contains channel id, related to the wire or grid which collected the charge, as well as the time-of-flight (46 bits) and collected charge (12 bits). All events which occur within a pre-defined coincidence window are received together. A majority of the events consist of one wire and a few grids. By combining coincidences between wires and grids, neutron events are reconstructed with a (x, y, z)-hit location, time-of-flight and collected charge. If more than one grid fired within coincidence, the grid with the most collected charge is used for the position reconstruction.

Method and analysis
This study concerns three properties of the Multi-Grid detector: neutron detection efficiency, internal neutron scattering, and time-and energy resolution. For the efficiency and resolution analysis, data from Multi-Grid Detector 1 are used, while for the scattering analysis data from both Detector 1 and 2 are needed. The method and analysis procedure is described below, starting with a reduction of the raw data by an event selection procedure.

Event selection
The Multi-Grid detector is designed for cold-to epithermal neutron detection. However, as the detector also has a certain low, but not negligible, gamma sensitivity [38], it is necessary to remove gamma events before proceeding with the analysis. This is done by studying the pulse height spectra (PHS), which shows the distribution of charge collected from events in the detector.
Gamma events have a clear signature in PHS. This is presented in figure 6a, where the PHS (x-axis) is plotted for each grid (y-axis). The y-axis is expressed in electronic channels, where Grid 80 is at the bottom and Grid 119 at the top (channel 0 to 79 are reserved for wires). The x-axis is expressed in 12 bit analog-to-digital converter (ADC) channels, ranging from 0 to 4095. The gamma events are concentrated in a distribution at the low ADC values, 100 to 500 ADC channels, while neutrons span the full range. Notably, neutrons from the direct beam are seen hitting only three grids at the lower part of the detector, as shown by the three horizontal red stripes.

Gamma events
Direct neutrons : Event selection on data from the Multi-Grid detector. In (a), the PHS (x-axis) are shown for all individual grids (y-axis). The numbering refers to the electronic channels related to the grids, where Grid 80 is at the bottom and Grid 119 at the top. In the figure, data from Detector 1 is used as an example. In (b), the exposed PCB connectors are shown. The center grid, Grid 99, is placed directly above the connectors.
There is an elevated noise level in the middle grid, Grid 99, where the events cover a larger range in the PHS. This middle grid sits directly on top of a junction between the two unshielded PCB connectors connected halfway up the detector, see figure 6b. The increased rate could, therefore, be due to crosstalk between the exposed connectors and the grid. Alternatively, a local physical offset would put the wire closer to the center grid, which would increase the gain there in a discrete step. As the nominal pitch between grids and wires is only 5 mm, a deviation on the fraction of mm would be sufficient to cause a noticeable effect.
For the purpose of the following analysis, the exact distribution of the gamma spectrum is not critical. Here, a constant ADC software threshold at 600 ADC channels is applied, such that the gamma events are rejected. However, for the scattering analysis, a more aggressive ADC cut is required to clean the data further, as the scattering signal is so low. There, a cut at 1200 ADC channel is used instead, which removes the remainder of the noise present in the middle grid.
In addition to the ADC-cuts, a multiplicity cut is performed. The multiplicity of an event denotes the number of wires and grids which fired within the coincidence window. As the conversion products have a finite range, there is a limit on how large multiplicities proper neutron events can have. Therefore, a study of the multiplicity can be used as an additional filter to remove gamma events and false coincidences. For this study, events with wire multiplicity 1 and grid multiplicity 1 to 5 is kept. The maximum grid multiplicity is larger than that for wires, as the released charge spread more easily to adjacent grids than adjacent wires.

Efficiency
Neutron detection efficiency is defined as the fraction of incident neutrons which are detected, i.e. the number of detected neutrons divided by the total number of incident neutrons. This is described in equation (1) where is the neutron detection efficiency, S detected is the sum of detected neutrons and S incident is the sum of incident neutrons. For the Multi-Grid detector, the quantity S detected M G is accessed from the number of counts in the data, while S incident M G is not measured directly. Instead, it is estimated from a separate measurement with the helium-3 tube. Using these two measurements, one with the Multi-Grid detector and one with the helium-3 tube, the neutron detection efficiency for the Multi-Grid detector is derived.
The incident neutrons on the Multi-Grid detector S incident M G , i.e. integrated absolute flux, is calculated according to equation (2), where S detected He−3 is the detected neutrons in the helium-3 tube and He−3 is the calculated efficiency of the helium-3 tube. BM M G is the integrated beam monitors counts over the full wavelength spectrum from the measurement with the Multi-Grid detector, and BM He−3 the corresponding number from the separate run with the helium-3 tube. The inverse of the helium-3 efficiency is used to scale S detected He−3 to approximate the total number of neutrons incident on the detector, S incident He−3 . The fraction of the beam monitor counts is then used to scale this value to make it comparable to the integrated flux on the Multi-Grid detector. That is, it accounts for the difference in flux during the two separate measurements.
Inserting equation (2) into (1), a formula for deriving the Multi-Grid efficiency as a function of the incident neutron wavelength λ n is thus written according to equation (3), where M G (λ n ) is the derived Multi-Grid efficiency. S detected M G (λ n ) and S detected He−3 (λ n ) are calculated by integrating counts above background for each peak from the Fermi-chopper in the wavelength spectra. This is presented in figure 7.  , an example of peaks at 3.6Å are presented for the Multi-Grid detector and helium-3 tube. In the plots, the Gaussian fit (dotted black), the background estimation (dashed black) and the narrow-(red) and wide (blue) integration intervals are also presented. In (c), integrated peak area is shown as a function of neutron wavelength. The peak area for the Multi-Grid detector (squares and diamonds) and helium-3 tube (triangles) is presented for the narrow peak interval (red) and wide peak interval (blue). The fraction of pile-up events encountered in the helium-3 read-out system (green) is plotted on the separate right-hand y-axis.
As the experimental setup does not produce well defined peaks, recall the parasitic peaks shown in figure 3c, there is an uncertainty on the correct way to integrate the peak area. This uncertainty is accounted for by introducing two intervals: one very narrow, ±σ, encompassing only the peak center, and one wide, encompassing the full peak, as well as any parasitic peaks. By using the narrow interval, effects from the parasitic peaks can be rejected. However, this also means that any differences in peak shape between the Multi-Grid detector and helium-3 tube are not accounted for. By using the two intervals, an estimate of this systematic uncertainty on the efficiency is obtained.
The peak areas are presented in figure 7c as a function of wavelength, where the areas from the narrow-(red) and wide (blue) are presented together. The highest peak neutron intensity [39] in the spectra, at around 2.5Å, is approximately 2 MHz cm −2 . The peak rate was estimated using data from the beam monitor, located a few meters upstream from the helium-3 tube.
He−3 (λ n ) is calculated to account for the absorption in the gas, the attenuation in the stainless steel tube, as well as the efficiency variation over the tube diameter. In figure 8a, the efficiency is plotted as a function of position along the tube diameter, together with measurement data from a previous measurement [40,41]. The data agree to within a few percentages, confirming the validity of the calculation. The offset is accounted for in later calculations as a systematic uncertainty. In figure 8b, the average efficiency across the tube diameter is shown as a function of neutron wavelength. Two curves are presented, one with the beam centered on the tube and one with a 5 mm offset. This is to account for the systematic uncertainty of beam alignment, which is taken into consideration in later calculations.  : Neutron detection efficiency of the helium-3 tube. In (a), the efficiency at 2.5Å is shown as a function of displacement from the tube center. The neutron beam width (red lines) is shown together with the calculation (black line) and data (black points), which was gathered during a previous measurement at ILL. In (b), the calculated efficiency is presented as a function of wavelength. Each value is averaged over the tube width hit by the beam, showing it centered (red) and at ±5 mm offset (blue).
In figure 9, the beam monitor data are plotted, corresponding to the separate measurements with the Multi-Grid detector and helium-3 tube, BM M G and BM He−3 , respectively. In figure 9a, the rates are shown, while in figure 9b the fractional rate, BM M G /BM He−3 , is presented. It is seen that incident flux during the two measurements were similar, as the fraction is only a few percentages below 1. It is also seen that the fraction is wavelength independent within a few percentages. The high uncertainty around 1Å is due to poor statistics. The derived Multi-Grid efficiency is presented in figure 10. In the plot, the derived efficiency for the two integration intervals (red and blue) are compared to the analytical prediction (black). The calculation for the analytical prediction includes the attenuation of neutrons in aluminum, the absorption probability in the 10 B 4 C-coating, as well as the escape probability of the conversion products from the coating. The incident neutrons are assumed to hit the front of the detector at a perpendicular angle. All calculations are based on derivations presented in [42,43,44].
Referring to figure 10, it is seen that for long wavelengths, 4 to 6Å, the derivedand calculated efficiency agree well within the uncertainties. For wavelengths shorter than 4Å (highlighted grey area), a strong deviation from the calculation is seen. This is due to the saturation of the helium-3 detector system. The peak flux is high enough, > 1 MHz, to cause multiple hits within the 1 µs shaping time of the read-out system. This is seen from the fraction of pile-up events in the tube (green), which follows the observed deviation. An additional reason is that the data transfer speed limit per channel, ≈ 10 6 events s −1 , is similar to the peak flux, which might cause loss of data. This could have been prevented by using a lower incident flux. However, as the scattering analysis requires the best possible signal-to-noise ratio, a high neutron flux was essential.
An attempt was made to account for the saturation in the helium-3 tube using information of the incident flux and shaping time. This is presented for the wide peak interval (green squares) and the narrow peak interval (orange diamonds). The procedure allows for four more data points to be within uncertainties. However, the correction is not strong enough for the majority of data points within the saturated region. This could be because the incident flux is sufficiently intense to introduce additional effects in the helium-3 tube, such as space charge effects, which further decreases the efficiency. As these additional effects are not accounted for in the correction, a deviation is still seen.
The saturation process is also the cause of the strong staggering effect between 2 and 3Å. In figure 7c, it is seen how the Multi-Grid detector (squares and diamonds) follow the intensity from Fermi-chopper, i.e. every other pulse is more intense, even where the flux is at the highest level. This is not the case for the helium-3 tube (triangles), which is flat between 2 and 3Å. As the oscillations between adjacent data points in the two detectors no longer match in this region, the fraction of the peak areas is not flat, as it is above 4Å. The saturation effect seen in the helium-3 detector is absent in the Multi-Grid detector. Pile-up fraction Unusable region Figure 10: Derived Multi-Grid efficiency (blue circles and red triangles) plotted against the calculated efficiency (black line). In plot, the derived efficiency where the helium-3 data has been corrected to account for the saturation in the helium-3 tube (green squares and orange diamonds) are also shown. The fraction of pile-up events encountered in the helium-3 read-out system (green crosses) is plotted on the separate right-hand y-axis. The grey region covers the unusable portion of the data, caused by saturation in the helium-3 detector system.

Internal neutron scattering
The main source of internal neutron scattering in the Multi-Grid detector is caused by the presence of aluminum [22]. Aluminum has a periodic crystal structure, like any other crystalline solid, which allows for coherent scattering of atoms within the same crystal lattice, as well as incoherent scattering from the individual atomic nuclei. The scattering cross-sections in aluminum are presented in figure 11a, where the coherentand incoherent components are presented separately. The cross-sections were generated using NCrystal [45]. It is seen that for wavelengths between 1 and approximately 4.7Å, coherent scattering is dominant. The cut-off wavelength at 4.7Å, where the coherent scattering drops to zero, indicates the maximum wavelength where diffraction occurs in the crystal structure, i.e. the maximum wavelength where the Bragg condition is fulfilled for the aluminum crystal lattice. For wavelengths longer than this, no coherent scattering occurs. An example of internal neutron scattering is presented in figure 11b.  Figure 11: Neutron scattering of aluminum. In (a), the cross-sections for coherent-(blue) and incoherent scattering (orange) are plotted as a function of neutron wavelength, together with the total scattering cross-section (red). The figure is generated using NCrystal [45]. In (b), an illustration of internal scattering is shown, where an incident neutron (blue) is scattered (red). The radial blades are highlighted in orange, which can be either coated or non coated with 10 B 4 C.
The impact of the scattered neutrons shows as an added time-dependent background in the detector. This is caused by an incorrect energy reconstruction of the scattered neutrons. The neutron energy, E n , is determined according to equation (4), where m n is the neutron mass, d is the source-to-detection distance, and tof is the corresponding time-of-flight. However, as the distance d is based on the detection voxel, this distance will be incorrect if the neutron is scattered before being detected. Moreover, as the neutron might acquire an additional flight time between scattering and detection, or a shortened flight time if the neutron gained energy after the scattering, this further distorts the energy reconstruction. Therefore, a scattered neutron has an energy reconstruction dependent on equation (5), where E n is the reconstructed energy for an internally scattered neutron, ±δT is the change in the flight time between scattering and detection, and ±δd is the change in assumed distance due to the incorrect voxel detection. Due to the large mass difference between aluminum nuclei and neutrons, most scattering can be considered elastic. Consequently, δT will predominantly be positive, corresponding to the extra flight time between scattering and detection. Therefore, internally scattered neutrons will predominately be reconstructed with an energy E n < E n .
To minimize the impact of scattered neutrons on the energy line shape, the additional flight distance between scattering and detection should be kept as short as possible. This reduces δd and δT in equation (5), closing in to the ideal case in equation (4). To facilitate this, 10 B 4 C-coating on the radial blades in the grids can be introduced (figure 11b, orange lines).
To investigate the effect of the radial coating, data from Detector 1 (non-coated radial blades) and Detector 2 (coated radial blades) are compared. As the neutron beam is highly collimated, recorded events from scattered neutrons can be separated from those from the direct beam. This is achieved by performing a geometrical cut, removing all events in the direct beam (figure 12a and 12b, short red stripes) and keeping the scattered neutrons (figure 12a and 12b, long green stripes). The volume outside the direct beam region is called the beam periphery. Note that there are scattered neutrons in the beam region as well, and that these are rejected in this approach. This is not an issue, however, as an absolute measure of scattering is not intended, only a comparison between Detector 1 and Detector 2.
To verify that the events seen at the beam periphery are indeed internally scattered neutrons, and not from a beam halo or a similar effect, the background data is used. This is presented in figure 12c and 12d, which shows data recorded when the direct beam was blocked with the helium-3 tube. The helium-3 tube, 25 mm in diameter, is wide enough to stop the direct beam. However, it does not have a sufficient solid angle coverage to make a prominent blocking of a potential halo effect. That the long green lines in figure 12a and 12b is only seen when the helium-3 tube is removed from the direct beam, demonstrates that the effect must originate after the helium-3 tube. As the only component after the helium-3 tube is the Multi-Grid detector, this shows that the lines originate from scattering in the detector.
Studying data with two different cuts, full data versus no direct beam, the wavelength distribution is acquired for Detector 1 and 2. This is presented in figure 13. The blue and green histograms are from the full data, while the red histograms are from data where the direct beam is cut. Overlaying the curves allows for a visualization of the magnitude of scattered neutrons and how it depends on wavelength. The black histograms are from the corresponding background measurements, which has an additional normalization based on the time-independent background level in the facility. This rate is seen in the flat region between 0 and 0.5Å, and depends on the overall activity in the vicinity of V20, which cannot be account for by the beam monitor data alone. It is noted that the background data has no time correlation with the Fermi- To study the energy line shape in detail, a peak at approximately 1.47Å from Detector 1 is used as an example, see figure 14. In figure 14a, histograms over energy are presented. Again, the blue histogram is from the full volume, the red histogram is from the beam periphery, and the black histogram is from the background measurement. Overlaying the curves allows for a clear visualization of how the scattered neutrons affect the shape of the peak. It is observed that the majority of scattered neutrons are reconstructed with a lower energy than the peak mean, resulting in a "shoulder" on the left side of the peak.
From the long green lines in figure 12a, it is seen that the maximum distance the scattered neutron travel within the detector is approximately 15 grids, corresponding to about 30 cm. To estimate where these scattered neutrons are reconstructed in the energy spectra, equation (5) is used. This is done by using the peak mean as an approximation of the incident neutron energy, while assuming elastic scattering and that δT is the dominant contribution to the distortion of the energy reconstruction. The colored vertical lines in figure 14a shows the result of this analysis, which overlaps well with the location of neutrons detected in the beam periphery (red). A comparison in the energy line shape with data from the helium-3 tube is presented in figure 14b. Four histograms are shown, three from the Multi-Grid detector and one from the helium-3 tube. The separate histograms from the Multi-Grid detector corresponds to the full data, beam periphery, and beam center. On the left side of the peak, it is seen that the helium-3 data overlap well with the data from the beam center of the Multi-Grid detector (compare orange and green). This further validates that the peak shoulder in the full data (blue) is due to scattered neutrons. Note that there is additional broadening in the data from the Multi-Grid detector, as it was further away from the Fermi-chopper than the helium-3 tube. This also affects the parastic peak on the right side of the peak center, which is present in both data sets but broader for the Multi-Grid detector.
To quantitatively compare Detector 1 and 2 in terms of internal scattering, a figure-of-merit, f om, is introduced. This is defined as the number of counts above background at a specified interval away from the peak center, divided by the peak area. The background is estimated to be locally flat, i.e. constant in energy, over the peak width. It is calculated on a peak-by-peak basis, based on the rate at the side of the peak. The reason the background measurements is not used for background subtraction 37 Figure 14: Effect on peak shape by internally scattered neutrons at 1.47Å in the Multi-Grid detector (Detector 1). Histograms from the full data in the Multi-Grid detector (blue), the beam periphery (red) and beam center (green) is seen, together with the background data (black) and helium-3 data (orange). In (a), the approximated location of scattered neutrons are shown as vertical lines. Each line corresponds to the energy reconstruction for scattered neutrons with a specific extra travel distance within the detector, as specified in the legend. The counts in the histograms are normalized to the accumulated beam monitors counts. In (b), the Multi-Grid detector is compared with the helium-3 tube. The helium-3 data has been artificially shifted along the energy-axis, such that it is aligned with the peak center of the Multi-Grid detector. This is to facilitate peak comparison. The counts are normalized to peak maximum.
is because of non-negligible systematic offsets. This is seen in figure 13a and 13b, where the background (black) does not follow the "bump" at 3Å equally well. The definition of the chosen f om is presented in equation (6), where peakarea is the peak area within ±5σ (using data from the full volume), E 1 and E 2 are the energy interval limits, counts is the number of counts in the beam periphery, and background is the flat background estimation. That is, the f om captures the background subtracted counts at the edge of the peak, as a fraction of the peak area. Consequently, a small f om is desirable, as this implies a low amount of scattered neutrons in the specified energy range. In words, the f om can be approximately stated as: f om = shoulder area peak area .
The limits E 1 and E 2 should be selected on a peak-by-peak basis, such that the same peak region is scanned for all peaks across the wavelength spectra. Unfortunately, this is not a trivial task, as the peak shape changes with wavelength. This also complicates the comparison in f om between different instruments, such as studies done at SEQUOIA and CNCS [11,35], as the peak shape is heavily dependent on the resolution of the chopper system. Here, the peak shoulder is split into approximately three equally sized regions, where the region limits are based on the standard deviation σ around the peak center. Each interval is 5σ wide, see vertical lines in figure 15 for an example at 1.47Å. The f om for Detector 1 and 2 are plotted in figure 16a, where each data point corresponds to a peak. The f om is shown for the three different integration ranges: [-5σ, 0σ], [-10σ, -5σ], and [-15σ, -10σ]. These relative intervals vary in absolute range in energy from peak to peak. The absolute energy interval, ∆E, varies from approximately ∆E = 5 · 10 −1 meV at 1Å, to ∆E = 5 · 10 −3 meV at 6Å. It is observed that for wavelengths above the aluminum cut-off wavelength at 4.7Å, f om is in most cases indistinguishable from zero. Furthermore, it is seen that for most wavelengths, Detector 1 (non-coated radial blades) has a larger f om than Detector 2 (coated radial blades). This is visualized more clearly in figure 16b, where the fractional f om is shown. Above 4.7Å, the uncertainties are large due to low statistics above the aluminum cut-off wavelength. It is noted that the effect resultant from coating the radial blades increase with distance from the peak center, and that the scattered neutrons are attenuated more strongly further away.  Figure 16: Comparison of the f om, as described in text, as a function of neutron wavelength, comparing Detector 1 (non-coated radial blades) and Detector 2 (coated radial blades). In (a), Detector 1 (squares and diamonds) and Detector 2 (triangles) are compared using three different integration ranges: [-5σ, 0σ] (blue), [-10σ, -5σ] (red), and [-15σ, -10σ] (green). In (b), the fractional f om, f om(Detector 2)/f om (Detector 1), is plotted. The three data sets (circles, triangles and diamonds) correspond to the three different integration intervals.
The average improvement of f om by coating the radial blades are presented for the three shoulder regions in table 2. 1 The results are based on data below 4Å, where the statistical uncertainties are acceptable. Note that these numbers are from a measurement with a collimated 14 × 60 mm 2 rectangular beam on the Multi-Grid detector, which is not generally the mode during normal operation. The reason a thin beam was nevertheless used was an experimental limitation, as the incident beam had to fit the helium-3 tube as well, which is much more narrow than the Multi-Grid detector. The direct beam of the current measurement hits the detector close to 90 • , within a single row, in between radial blades. However, during normal operation, neutrons will in many cases hit a much larger portion of the detector surface area (with the important exception of Bragg diffraction from crystals, which can be a few cm 2 on the detector surface), and typically at angles with a small divergence around 90 • . This will influence the effect of coating the radial blades, as the absorption of neutrons depends on two factors: incident angle on the radial blades, and the number of radial blades crossed. Table 2: Average improvement of f om between 1 and 4Å by coating the radial blades. The results are presented separately for the three shoulder regions.

Shoulder region Average improvement
Incident neutrons hitting the radial blades at an angle are absorbed earlier. This increases the efficiency and decreases interaction opportunities in aluminum. This effect has been predicted by simulations at this geometry [23]. As the incident neutron beam was kept within a single row, i.e. without crossing the radial blades when incident, this effect might not be as prominent in the current measurement.
Concerning the number of radial blades crossed, this depends on how close to the detector edge the neutron is scattered, see figure 17. If the neutron scatters from the center, the maximum extra travel path is identical to the right and left. This extra travel path determines the impact on the energy line shape, recalling equation (5). However, if the neutron is scattered close to the edge, there is an asymmetry in the effect on the line shape from neutrons scattering inwards and outwards. For neutrons scattering inwards, the maximum effect is greatest, as the maximum extra travel distance is the longest. These events are also, incidentally, best attenuated by coating the radial blades, as the neutrons need to cross up to twice as many blades when scattering inwards from the side compared to scattering from the center. All these edge scatterings are not accounted for when only the center is illuminated. Consequently, a stronger attenuation of scattered neutrons by coating the radial blades is expected during normal operation.
An option to further attenuate scattered neutrons is to use the gap between adjacent grid columns and insert good quality neutron shielding there, i.e. quasi black. This reduces the maximum scattering distance from 11 to 3 cells, hence further reducing the impact on the energy line shape. This might be a more cost-effective solution, compared to further increasing the thickness of the radial coating, to get an improvement in line shape.
From measurements at SEQUOIA [35], preliminary results support the predicted greater advantage of using the coated radial blades during normal operation. At SEQUOIA, the same Multi-Grid detector prototypes as the ones used at V20 were compared in normal operation, i.e. neutrons scattered of a sample and hitting a large portion of the detector surface. From the results, the improvement is seen to be approximately 38 ± 10 %, averaged over the available wavelength spectra, 1 to 3Å. The improvement is the largest around 3Å, reaching a maximum of 55 %. Note that the improvement obtained from the SEQUOIA measurement is not directly comparable to the improvement stated in this paper, as the results are based on a different figureof-merit definition. First, it is not possible to separate scattered neutrons from direct neutrons in the SEQUOIA data, which is done with V20 data using a geometrical cut, because neutrons isotropically hit the full detector surface. Therefore, there will inevitably be neutron events from the direct beam included in the shoulder in the SEQUOIA data.
Second, the result from SEQUOIA is based on data over the full shoulder, not just small sections of the shoulder as in the V20 numbers. Considering an approximately equivalent range at V20, [-15σ, -5σ], over the same wavelength range, results in a 23 ± 8 % improvement. This is lower than the 38 ± 10 % obtained from the SEQUOIA data, which indicates that the impact of the radial blades will be stronger during normal operation.
All the presented percentage errors should be interpreted as variations around the mean, not statistical uncertainties. It is not predicted that the radial blades will have the same effect across all wavelengths. Thus, a mean with a low variance is not expected. This is because the effect of the radial blades depends on the number of radial blades crossed, which depends on the scattering angle of the neutron. As the scattering angle is wavelength dependent [45], the effect vary as a function of wavelength, as was shown in figure 16b. The stated standard deviations around the mean have the intention to capture the magnitude of this variation over the measured wavelengths.

Time-and energy resolution
The observed energy resolution depends on two main components: the resolution of the detector and the resolution of the chopper system. The observed peak shape is a convolution of the chopper pulse shape and the detector response shape, according to equation (7), where f observed is the observed energy distribution, f chopper is the energy distribution from the chopper system, and f detector is the detector energy response function.
To study the detector resolution component, it is crucial to keep the chopper distribution f chopper as sharp as possible. The width of this distribution depends on two main parameters: the chopper window opening time and the chopper-to-detector distance. The opening time determines the initial width of the neutron pulse, while the chopper-to-detector distance determines the amount of time broadening before the pulse reaches the detector. For this measurement, the small Fermi-chopper allows for a short opening window, while keeping the chopper-to-detector distance to a few tens of cm. This results in a highly resolved chopper pulse, which allows for a study of the detector response function.
As the helium-3 tube and the Multi-Grid detector are at different distances from the Fermi-chopper, recall figure 2, the width in energy spectra from Multi-Grid data are corrected to account for this offset. In figure 18a, the tof spectra are shown individually for each of the twenty wire layers in the Multi-Grid detector, together with the tof from the helium-3 tube. In figure 18b, the energy spectra corresponding to these tof -values are presented. In the figure, data from all twenty layers from the Multi-Grid detector are seen to be reconstructed at the same position, overlapping with the helium-3 data. However, there is an additional broadening on the Multi-Grid detector, caused by the additional distance from the Fermi-chopper.
To account for this, the peak width in energy is calculated using a Gaussian fit for each layer, see figure 18c. 2 This is to follow how the peak width, defined as the full width at half maximum (FWHM), depends on the distance from the Fermi-chopper. In figure 18d, the FWHM for each of the layers in the Multi-Grid detector, as well as the FWHM of the helium-3 tube, are plotted as a function of the distance from the Fermichopper. A linear fit to the data points from the Multi-Grid detector is also shown, together with the fit uncertainties. The interpolation takes into account the widening of the pulse with distance, and allows the Multi-Grid detector to be compared to the helium-3 tube. That is, the width the Multi-Grid is interpolated to the value it would have had, at the same distance from the Fermi-chopper as the helium-3 tube. Here the procedure is shown for a peak at 2.8Å as an example, but the same analysis is done for the peaks at the remaining wavelengths.  Figure 18: Correction for pulse broadening with distance, using 2.8Å neutrons as an example. In (a), the tof from the twenty individual layers in the Multi-Grid detector, from front to back, is presented together with the helium-3 tube (red). In (b), the corresponding energy spectra for the Multi-Grid detector (blue) and helium-3 tube (red) are shown. The histograms in (a) and (b) are normalized to peak maximum, to facilitate peak comparison. In (c), the widening in energy is presented for each layer. The distribution from each layer is fitted with a Gaussian (black) and the FWHM is extracted from the fit parameters. In (d), the FWHM from each Multi-Grid layer (blue) is shown together with the FWHM of the helium-3 tube (red), as a function of distance from the Fermi-chopper. A linear fit (solid blue line) from the Multi-Grid points is also presented together with the fit uncertainties (two dotted blue lines), demonstrating how the chopper pulse widens with distance.
Using the interpolated pulse width of the Multi-Grid detector, the detector energy resolution can be compared to that of the helium-3 tube. This is shown in figure 19, where the FWHM of the two detectors are presented as a function of neutron wavelength.
It can be seen that for the two data points corresponding to the shortest wavelengths measured, the Multi-Grid detector has a slightly higher energy resolution than that of the helium-3 tube. For the remaining wavelengths, conversely, the helium-3 tube has a finer energy resolution. The difference between the detectors is more prominent at longer wavelengths, where the chopper resolution is finest. The observed difference relates to the variation in time resolution between the two detectors technologies. The crucial difference between the detectors is the neutron absorption time distribution, which depends on the conversion location uncertainty, see blue and red lines in figure 19. These curves show how the uncertainty in conversion location translates to a corresponding uncertainty in energy. This uncertainty can vary more in the helium-3 tube than that of the Multi-Grid detector, as the absorption reaction can take place anywhere in the gas along the 2.5 cm depth of the helium-3 tube. For the Multi-Grid detector, however, the absorption only takes place at discrete intervals of 0.95 cm (1 cm pitch − 2 · 0.025 cm aluminum thickness), where the solid conversion layers are located. The two curves are always beneath the data points as they do not account for the remaining time uncertainties in the measurement, such as chopper pulse broadening.
In figure 20, the conversion location uncertainties for the helium-3 tube and Multi-Grid detector are compared. The uncertainty in the helium-3 tube was acquired using a one-dimensional calculation of the neutron absorption as a function of depth, which was done for a series of different incident wavelengths. As an estimate of the detection location uncertainty, the depth where 95% of the detected neutrons are converted is used. Only the range is considered, as this is estimated to be the dominant factor of the neutron detection distribution. However, factors such as tube shape and the conversion products ranges also contribute to the distribution. Using a high percentage, 95%, is an attempt to account for these missing factors.
For the two shortest wavelengths measured, the Multi-Grid detector has the advantage, as a non-trivial amount of neutrons travel further than 0.95 cm into the helium-3 tube before absorption. However, for longer wavelengths, the helium-3 tube performs better, as the majority of neutrons are absorbed within less than 0.95 cm from the surface. This is in contrast with the Multi-Grid detector, which still has the fixed 0.95 cm timing difference. This is further illustrated in figure 21, where the energy-distribution for a peak at 4.3Å is presented together with the corresponding tof -distribution. In figure 21a, a double peak is seen for the Multi-Grid detector (blue), while the helium-3 tube (red) has a single peak. The double peak is a consequence of the discrete 0.95 cm distance between consecutive conversion films. This is more clearly visualized in figure 21b, where tof -data from each wire layer are plotted separately. It is seen that each wire layer has a double peak. This is because each wire layer is adjacent to two converter films, where neutrons can be absorbed in either one. The distance between the peaks is approximately 15 µs, which, for a 4.3Å neutron, translates to 0.91 cm. This is close to the expected 0.95 cm timing difference between adjacent converter films.
It is worth noting, that the better the peaks are separated, the more information is gained concerning which coating surface the absorption took place. This could potentially be used to increase the spatial resolution of the Multi-Grid detector down to the specific surface the conversion took place, providing the peaks are sufficiently well separated. Although this investigation is beyond the scope of this work, it might be worth exploring in the future. From the analysis a clear double peak is seen from the Multi-Grid detector for the longest wavelengths measured. However, in practical applications of the Multi-Grid detector at an instrument, such as at CNCS and SEQUOIA, this effect has not been observable. This is due to two primary reasons.
The first is additional pulse broadening, which is due to the approximately 1 order of magnitude longer chopper-to-detector distance in a typical instrument. A comparison can be made between the current setup, which has a few tens of centimeters chopperto-detector distance, with the CNCS and SEQUOIA instruments, where the equivalent distance is a few meters. As the pulse broadening is proportional to the distance, a corresponding one order of magnitude more broadening is seen at CNCS and SEQUOIA for the same wavelength range measured here [11,35]. The effect of the double-sided coating is below the instrument resolution, and no difference in energy resolution is seen between the helium-3 tube and Multi-Grid detector.
The second reason for the missing double structure in the data from the CNCS and SEQUOIA instruments, is the longer flight time between Fermi-chopper and detector. If the neutron time-of-flight is short, the energy reconstruction will be more sensitive to deviations around this value. Recalling equation 5, the sensitivity depends on the relation between the time-of-flight tof and perturbation ±δT . By having a longer timeof-flight, such as at a typical instrument, the energy reconstruction will be less sensitive to the additional time δT between adjacent converter layers, which will be very small in comparison to the full tof . Hence, the timing-difference between the two layers will be less pronounced.

Conclusions
Using a series of ∼ µs short neutron pulses, chopped by a table-top prototype Fermichopper, it is possible to investigate subtle time-and energy resolved effects in the boron-10 based and helium-3 based thermal neutron detectors, reaching the resolution of a few µeV. Properties which can be studied include internal neutron scattering in the Multi-Grid detector, as well as a comparison in energy resolution with the helium-3 tube. In addition to this, the neutron detection efficiency of the Multi-Grid detector can be derived using a complementary measurement with the helium-3 tube.
The derived efficiency of the Multi-Grid detector is observed to match the theoretical prediction for wavelengths above 4Å, and be close at 1.3Å. For the remaining wavelengths, a strong discrepancy is seen, caused by saturation in the helium-3 detector system, as the detector and read-out system was unable to cope with the peak flux at these wavelengths. The Multi-Grid detector shows no sign of saturation. This reemphasizes the need for neutron detector systems with high count rate capabilities.
Concerning internal neutron scattering in the Multi-Grid detector, it is seen to be primarily caused by neutron scattering in the aluminum structure within the Multi-Grid detector. Above 1Å, coherent scattering is the dominant component until approximately 4.7Å, where it drops to zero. This is reflected by the results, which show that above the aluminum cut-off wavelength at 4.7Å, the scattering is below the time-independent background level.
A detailed study on how the internally scattered neutrons affect the line shape was done using a geometrical cut, discarding the region where the beam hits the detector. This removes all events from directly absorbed neutrons, while keeping events from neutrons that are only absorbed after first being internally scattered. This allows an investigation of which part of the peak is due to scattering.
By studying data from Detector 1 (non-coated radial blades) and Detector 2 (coated radial blades), the net effect on the line shape is compared. Between 1 and 4Å, the average improvement induced by coating the radial blades is seen to increase with distance from the peak center, starting at 7 ± 6 % between [-5σ, 0σ], and reaching 31 ± 8 % at [-15σ, -10σ]. This result is obtained from a direct beam on a small part of the detector. When a larger portion of the detector surface is illuminated, as during normal operation, the effect of the radial blades is expected to be stronger. This is supported by preliminary results from the SEQUOIA measurements.
The energy resolution of the Multi-Grid detector and helium-3 tube is observed to broadly match, within 5 µeV, over the measured wavelengths. For the shortest wavelengths, below 1.5Å, the Multi-Grid detector has the advantage, while above 1.5Å, the helium-3 tube performs better. This results from the fact that, for the helium-3 tube, neutrons can be converted anywhere in the gas along the 2.5 cm depth of the tube, while the absorption in the Multi-Grid detector only occurs at discrete intervals of 0.95 cm. This results in a difference in timing resolution. Due to this, at longer wavelengths, the obtained energy spectra from the Multi-Grid detector have a double peak, where each peak originates from one of the two conversion films adjacent to the same wire layer. This effect becomes more evident for longer wavelengths, as the additional path difference is more important for slow neutrons than for fast ones. However, in practical applications of the Multi-Grid detector at an instrument, such as at CNCS and SEQUOIA, this effect has not been observable. This is due to the longer flight path and additional pulse broadening at these facilities, which causes the 0.95 cm timing difference to be below instrument resolution. Finally, note that the measured resolutions are a function of the detector specifications. The energy resolution of the helium-3 tube can be varied by choosing a different gas pressure in the tube, and, to a lesser extent, by changing the geometry. Conversely, the resolution of the Multi-Grid detector can be adjusted by varying the distance between layers. It can also be changed by choosing a single-sided coating approach, whereupon the uncertainty concerning which coating the conversion took place would vanish. That is, the double peak observed for long wavelengths would disappear, as it is unambiguous in which conversion layer the reaction occurred. However, a consequence of this would be either to accept a reduced efficiency, or a doubling of the detector depth. In both scenarios, the internal scattering would have a larger impact on the line shape, as the ratio between converter material and aluminum within the detector would be smaller. Thus, there would be less converter material to absorb the scattered neutrons. This results in a design trade-off between energy resolution, background reduction, and efficiency.