HEKATE—A novel grazing incidence neutron scattering concept for the European Spallation Source

Structure and magnetism at surfaces and buried interfaces on the nanoscale can only be accessed by few techniques, one of which is grazing incidence neutron scattering. While the technique has its strongest limitation in a low signal and large background, due to the low scattering probability and need for high resolution, it can be expected that the high intensity of the European Spallation Source in Lund, Sweden, will make many more such studies possible, warranting a dedicated beamline for this technique. We present an instrument concept, Highly Extended K range And Tunable Experiment (HEKATE), for surface scattering that combines the advantages of two Selene neutron guides with unique capabilities of spatially separated distinct wavelength frames. With this combination, it is not only possible to measure large specular reflectometry ranges, even on free liquid surfaces, but also to use two independent incident beams with tunable sizes and resolutions that can be optimized for the specifics of the investigated samples. Further the instrument guide geometry is tuned for reduction of high energy particle background and only uses low to moderate supermirror coatings for high reliability and affordable cost. ➞ 2018 Author(s). All article content, except where otherwise noted, is licensed


I. INTRODUCTION
2][3][4][5][6] The studied systems cover a broad range of scientific fields from self-organized nanoparticles over surface structures in free floating polymer solutions to magnetic structures like skyrmions.Using the beam under a grazing angle allows not only to enhance scattering from surface-near structures due to dynamic effects, but under an angle slightly below the critical edge of total reflection, it is even possible to limit the penetration depth of the neutrons deterministically and thus only probe surface-near structures.In comparison to the similar x-ray technique (Gracing Incidence Small Angle X-ray Scattering, GISAXS), these investigations with neutrons have been limited to quite a low number of experiments for a variety of reasons.GISANS scattering is very weak, and due to the small angle of incidence, the projected sample size in the reflection direction is very small, while the in-plane size of the sample is several mm.Due to dynamic effects close to the critical edge of total reflection, the modeling of the measured intensity is challenging, user friendly software was long missing, and long computing times were necessary, especially when considering the relaxed resolution compared to x-rays.Recently, the BornAgain 7 software and alternative models 8 have been developed including modified kinematic theories, 9 to address the need for such simulations.a) Electronic mail: artur.glavic@psi.chFor a more detailed discussion of the GISANS technique, covered science areas, and challenges, see the review article by Müller-Buschbaum. 1 Traditionally these experiments were either performed on neutron reflectometers or Small Angle Neutron Scattering (SANS) instruments, each of which has its advantages and limitations when GISANS is investigated.While reflectometers are optimized to precisely control the incident angle of the neutron beam with well-defined divergence and cover large reflection angles, their collimation and sample-to-detector distances are typically relatively small.To achieve sufficient GISANS resolution, this often means that the beam size has to be reduced to only a fraction of the sample size, leading to very low statistics.These instruments also tend to provide better absolute wavelength resolution than necessary for low q measurements performed with longer wavelength as in GISANS, further reducing incident intensity ( q is the neutron scattering vector with q z component normal to the surface).SANS machines, on the other hand, often lack capabilities to properly orient the samples and cannot move the detector far out of the direct beam direction, thus requiring a reduced detector distance with lower resolution to cover larger reflection angles.In addition, it is difficult to relax the resolution of the instrument just in one direction and the inclination of the beam to access free liquid surfaces is mostly absent for SANS machines.For these reasons, the OS and specular reflectivity capabilities lack behind specialized reflectometers.In timeof-flight (TOF) instruments, the long collimation length of SANS also leads to a limitation in bandwidth while achieving wavelength resolutions better than necessary.
The European Spallation Source (ESS) is the next generation neutron facility currently being built in Lund, Sweden.
While the first set of instruments is already in detailed design, there are plans to select additional instrument concepts for a second suite of beamlines, including a dedicated surface scattering instrument.Here we present a concept for the latter, adapted to the spatial constraints and the neutron beam characteristics at the ESS.This instrument study, HEKATE for Highly Extended K range And Tunable Experiment, is optimized for grazing incidence scattering and reflectometry from a wide range of materials, including free liquid surfaces.In conjunction with most ESS instruments, it is named after the moon goddess of hidden wisdom.The neutron guide solution we have chosen allows to access a very large range of the surface normal component of the incident wave vector (k i,z = 2π λ sin α i ) even for free liquid surfaces, covering specular reflectivity from q z = 10 ☞3 to 0.3 Å ☞1 .GISANS resolution is obtained by focusing on the detector with an achievable resolution of down to q y min = 10 −4 Å −1 while keeping the sourceto-detector distance as little as 25.5 m to maintain a bandwidth >10 Å.Such good q y resolution enables the measurement of GISANS signals into the range accessible in OS as q x , thus allowing a coverage of various length scales.For a detailed discussion of the various aspects of the instrument resolution function, see the supplementary material.Flexible optics and bandwidth selection allow a precise tuning of the incident beam to the required experimental conditions, and low background is facilitated by a small monolith opening, early limitation of the beam size and two line-of-sight avoidances in the vertical and horizontal directions.With these features, the instrument would make optimal use of the high brilliance ESS long pulse source and allow up to now impossible grazing incidence experiments.

II. INSTRUMENT LAYOUT
While the general concept of HEKATE could be applied to other pulsed neutron sources, the instrument layout was defined based on the ESS plant geometry and shielding concept.A conceptual sketch of the neutron guides and key components is shown in Fig. 1.The basic idea behind the guide concept is to extract two separate beams from the cold moderator with one (here labeled a) covering a large divergence range for small incident angles and longer wavelength and the second (b) with smaller divergence but down to short neutron wavelength to reach extended k i,z values.The horizontal geometries for both beams are the same and allow focusing either on the sample for reflectometry or on the detector for GISANS.Due to spatial separation of the beam divergence, one double disc chopper may then be used to select different wavelength bands for individual incident beam angles for simultaneous coverage of a very large continuous k i,z range.
In the vertical direction, beam (a) gets extracted from the moderator (1, the numbers refer to the labeling in Fig. 1) under 1.2 • downward inclination using an elliptical feeder (2a) that focuses the beam to a virtual source (VS) slit at 7.5 m distance (3a).Following the VS, a Selene type neutron guide 10 (5a, 6a) transports the neutrons to the sample surface at 22.5 m (10), covering 0 • -2 • incident angle on a horizontal sample surface.For beam (b), the extraction uses a parabolic feeder (2b) with focus on the VS at 6.5 m, followed by a smaller Selene guide (5b, 6b) to hit the sample with 2.5 • -3.5 • incident angle.Before the entrance or between the two ellipses of the Selene guides, logarithmic spiral shaped transmission mirrors can be added for polarization and for frame overlap suppression as will be done for the Estia instrument. 11Both guides use the same geometry in the horizontal plane.A parabolic feeder (2t) focuses on an extended VS at 7.0 m (3t), and a larger single elliptical neutron guide (7) transports the beam to the sample position (10r).
With the chosen geometry and divergence profile, the extraction angle from the source is between ☞3.2 • and ☞0.6 boundaries.For the geometries and beam properties mentioned earlier, both feeders (2a, 2b) can be coated with m = 2.5 supermirrors (m describes the critical edge of total reflection relative to that of nickel).Beam (a) then requires a m = 2.5 coating on the side wall (7) and m = 3.5 on the Selene guide (5a, 6a).The values for beam (b) are m = 4 on all surfaces (5b, 6b, 7).While we did not perform a systematic optimization of the divergence and coating parameters that might improve the instrument at a later stage, the resulting performance is already very promising as shown below.

A. Chopper
Coupled to the selection of optics is the usage of the wavelength bands.A double disc chopper as shown in Fig. 2 with 14 Hz (source frequency) is placed between the two absorbers of the horizontal VS at 7 m (4).At this position, the beams (a) and (b) are separated so that the upper chopper disc allows for a switching between both.This is illustrated by the blue and red openings, which correspond to the non-overlapping time frames 1 and 3 in the time position diagram with the detector at 25.5 m and 2.75 Å minimal wavelength, shown in Fig. 3. Due to the distance of the chopper to the VS of beam (a), its angle of incidence on the sample is correlated to the vertical position within the beam cross section.Thus shading of the upper or lower part of beam (a), respectively, results in spatially separated regions on the detector being illuminated.Shifting the first chopper disc down with respect to the 2-frame mode has the result that the upper part of beam (a) experiences a different chopper opening (green) and forms frame 2 which partially overlaps frames 1 and 3 in time but not spatially on the detector.Height adjustment of the upper disc can be used to redistribute flux and k i,z coverage between frames 1 and 2 and this way to optimize the counting statistics over a broad q z range.
With the lower chopper stopped, each source pulse results in neutrons arriving on the detector spread over two pulse periods.If OS or incoherent scattering leads to a spatial cross talk on the detector, the lower chopper is used to select either only frame 1 or 3 (not shown), or both with reduced frame width (as illustrated in Fig. 4).If a better, constant relative resolution is required, an additional set of wavelength frame multiplication (WFM) choppers ( 13) can be introduced to achieve 1% ∆λ/λ, as used in other ESS instruments. 12Due to the small beam size in the vertical direction (<1 mm), this could be achieved with relatively small and slow choppers.
The various operation modes adapted to the needs of gracing angle scattering, reflectometry, different q ranges, and resolutions are discussed below together with the measurement schemes.

B. Detector features
The source-to-detector distance of 25.5 m together with the FWHM time resolution of ∆t = 3 ms of the source pulse leads to a wavelength resolution of ∆λ = 0.465 Å.
FIG. 2. Possible setup of a double disc chopper that allows the separate wavelength bands with the three beams passing in the center (color code as in Fig. 1).Shifting up the upper chopper can be used to switch from three (left sketch) to two (right sketch) distinct bands, and activation of the lower chopper switches from the specular reflectivity to grazing incidence mode with only one beam being active at a given time.For a typical wavelength band used for GISANS experiments (7.0-17.5 Å), this would correspond to a relative q resolution between ∆q/q = 6.6% and 2.7%.To match the instrument performance, the detector (11) would likely use the same multiblade technology 13 that is being developed for the reflectometers Estia and FREIA at ESS.This technology allows very high counting rates and resolutions of down to 1.0 × 0.5 mm 2 , and the thin entrance window guarantees a low broadening of the spatial detector response.With the proposed geometry a detector of 700 × 700 mm 2 would be sufficient to cover 0.1 Å ☞1 at 7 Å wavelength with the direct beam in the center.Current estimates indicate that this type of detector could even take the full direct beam, eliminating the necessity for a beamstop.If this is not feasible, a set of up to 4 separate beamstops should be available to cover direct and reflected beams from two beam paths.FIG. 4. TOF diagram for the grazing incidence scattering mode with two separate beams.Only parts of each band reach the detector (white overlay), with the longer red band being directed under higher angle to the sample surface leading to similar k i as for the shorter wavelengths.In this example, the bands are selected to switch at the prompt pulse time and cover the same k i range.

C. Focusing GISANS
For GISANS measurements, a hyperbolic mirror (8g) is inserted before the sample that shifts the focal distance in the sample plane to the detector at 25.5 m.In this configuration, the horizontal experimental resolution can directly be chosen using the size of the VS while homogeneously illuminating samples up to 20 mm width.Such a mirror could also be used as a second option to spin-polarize the neutrons for magnetic measurements.
In this mode, the second chopper is operated in-phase with either frame 1 or 3 and a vertical slit after the neutron guides define the incident angle and resolution.For free liquids, beam (a) with frame 3 would be used to be able to control the incident angle; experiments on solids would select the flux center of beam (b) with frame 1 for better performance on the short wavelength side of the spectrum.
For modern supermirrors, roughness is so low that OS contributions, which could increase background, can be neglected.Due to relatively small distances involved, the mirror alignment for both the neutron guide and the hyperbolic mirror are at least one order of magnitude less accurate than for the Estia instrument.

D. Duplex GISANS
By tilting one element within the second ellipse of the second guide (6b), one can couple out a part of the beam and redirect it to the sample with a small supermirror reflector, reaching very low angles of incidence (yellow beam and inset in side view, 12g).Activating, in addition, both choppers allow for a special mode of operation where two collimated beams hit the sample, beam (a) under slightly higher angle than beam (b).Setting the phases of the two choppers in the right ratio, only one of these beams will be open at a given time with the shorter wavelengths at lower incident angle than the longer ones.Figure 4 illustrates a possible setting where the small angle beam would cover 7-11 Å and the higher angle 11.5-17.7 Å, switching at the time of the prompt pulse.By selecting the ratio between both angles to be 11.5/7 = 1.6, both beams cover the same k i,z range and allow measurements below the critical edge of total reflection either with twice the incident intensity or with two different GISANS resolutions.Shifting the phases of the choppers allows us to freely distribute the band between both beams and thus a flexible tuning of two distinct incident beams, and the spatial separation of the virtual source slits adds the ability to tune the beam size and resolution individually.
Example applications include the mentioned doubling of incident wave vector measurement (equal distributed relative bandwidth and inverse incident angle), addition of higher q range measurement for selected incident wavevectors [short band for beam (b) with a small incident angle and extended band for beam (a)], or measurements with two regions of interest [e.g., beam (b) at higher angle concentrates on a feature near a Bragg-sheet while beam (a) probes the critical edge].

E. Reflectometry
A surface scattering instrument should also enable specular reflectometry so that all relevant measurements can be made to analyze the OS and GISANS data.However, the requirements concerning resolution, q z , range, and counting time for these specular measurements are relaxed compared to dedicated reflectometers, and these can thus be carried out with shorter wavelengths.(This is in contrast to penetration depth limited GISANS measurements close to the critical edge, since these require high resolution for the incident angle.) For time dependent studies on samples with low intrinsic background HEKATE can be used to measure focusing reflectometry as the Estia instrument, thus making use of the full incident divergence of 2 • + 1 • and covering a continuous band of specular reflectivity from 10 ☞3 Å ☞1 to 0.3 Å ☞1 with a wavelength band distribution as shown in Fig. 3, i.e., the lower chopper is stopped and all time frames are used with maximum divergence.
It is not possible to use the focusing mode for specular reflectivity from all samples.If the sample surface is bent or the sample itself produces considerable background from OS or incoherent scattering, the use of a collimated beam is required.This conventional TOF mode can be easily realized by reducing the divergence with a slit behind the guide exit and the selection of only one time frame by activating the lower chopper disc.This is also the mode used to measure OS.
In most cases, however, it would be advantageous to use a grid of selected openings to measure multiple angles simultaneously instead of sequential.While this would intuitively lead to increased background, the proper selection of openings for each beam, matched to the wavelength resolution in the HEKATE instrument, does not lead to this issue.For the smallest angles, the grid openings can be as small as 0.1 mm, but only for the part of the beam that transports wavelengths above 10 Å, allowing the use of a very thin Cd absorber.The focusing characteristics of the Selene guide will allow HEKATE to avoid issues with this approach, like beam cross talk, that have been found for such applications on conventional reflectometers.This multi-beam-focusing mode is presented in more detail in Sec.III.

F. Background
Background is a very important issue in grazing incidence experiments, especially for long pulse spallation sources, which tend to produce a large number of high energy particles.There are several features that allow HEKATE to keep this source of background as low as possible.First, the opening in the first shielding layer (2), the steel monolith that reaches from 2 m to 5.5 m, is only 18 mm wide and 22 mm + 18 mm high, leading to an open beam area of 420 mm 2 , which is 5% of that of the FREIA or a single beam of the Estia instrument.The vertical guides each lose the line of sight twice, the first time within the bunker wall shielding and the second time after the second elliptical mirror of the Selene guide.Horizontally the elliptic guide leads to a single line of sight avoidance, and with the hyperbolic reflector, the line of sight is lost a second time (if shielding is added behind the reflector properly).Due to the aberration corrected projection capabilities of the Selene neutron guides, the beam can already be reduced to the necessary size within the neutron bunker and the spatial separation of horizontal and vertical VS slits allows the horizontal ones to be simple absorber blades that could be equipped with heavy collimation shielding, acting as a pinhole for high energy neutrons.

III. MODELED PERFORMANCE
We have performed Monte Carlo neutron ray tracing simulations for the HEKATE instrument using the McStas 2.4.1 software. 14,15All simulations use the current 2016 ESS cold moderator model at beamport E08 and include the influence of gravity.First, the overall neutron guide transmission has been studied by placing a wavelength dependent monitor at the source and sample position while restricting the detection size to 10 × 2 mm 2 and the divergence to the acceptance range of the horizontal and vertical guides.Figure 5 shows the resulting neutron intensity as a function of wavelength for each guide as well as the brilliance transfer (inset) that was calculated by dividing the sample monitor by the source monitor.Guide (a) with the higher divergence angle and stronger "bending" of the beam does not transmit wavelengths much shorter than 5.5 Å but due to the larger total divergence and opening has about 2 times the transmission for long wavelengths and does not suffer from losses for very long wavelengths due to gravity.For guide (b), on the other hand, the lower reflection angle allows very short wavelengths to pass through the guide system for a flux peak at ≈2.6 Å.The spectrum of guide (b) is therefore optimal for reflectometry to very high q z values.
We have performed simulations with the focusing reflectivity mode (comparable to Estia) that demonstrate the good performance for specular measurements over a large q z -range with good statistics in a few seconds.Up to q z = 0.24 Å ☞1 , even 1 s measurements are sufficient to stay below 20% error.For more details, see the supplementary material.FIG. 5. Performance of the two neutron guides used in HEKATE simulated as neutron flux on an area of 10 × 2 mm 2 around the sample position.At the sample position on an area of 10 × 2 mm 2 .The green simulation in the GISANS mode uses a resolution that yields a 30 mm × 30 mm beam size on the detector.The inset shows the brilliance transfer for the same area and the full guide divergence of 1.4 • × 2.0 • (a) or 1.4 • × 1.0 • (b).While guide (a) transmits larger divergence and is not affected by gravity very much, the guide (b) has good performance for short wavelength, with its peak intensity at just 2.6 Å.
For samples with high intrinsic background, e.g., from incoherent scattering of hydrogen in water, a different approach has to be taken.The simulation shown in Fig. 6 was calculated using 8 beams directed to the sample, which are offset in time and in angle relative to each other: the first 6 using long (13.5-24.0Å), the 7th using intermediate (7.5-18.0Å), and the final beam using short wavelength (2.75-11.25 Å).Beam divergence is matched to wavelength resolution by using ∆θ/θ = 2%, 3%, and 10% for long, medium, and short wavelength bands, respectively.(In practice, the smallest openings, which are below 0.1 mm, will not be reached.Increasing these to the minimum achievable size would not change the impact on the instrument background.)Due to the small size of the openings for low incident angles, the number of Monte Carlo trajectories reaching the detector is very low, leading to increased noise in the simulated intensity values.For these beams, the angles were selected reach a large q z range coverage with sufficient overlap.As can be seen, the covered q z range is almost as large as in the full focusing mode, covering 2 × 10 ☞3 Å ☞1 to 0.29 Å ☞1 .For the q z range possibly affected by the sample background, however, only the highest angle is relevant.If one compares the total number of neutrons that hit the sample surface from the individual beams, beams 1-6 contain 2.2 × 10 6 s ☞1 , the 7th contains 6.0 × 10 6 s ☞1 , and the 8th contains 2.7 × 10 8 s ☞1 .Thus adding beams 1-7 to the necessary measurement for higher q z values increases the background only by some 3%.While no attenuation due to neutron windows and air scattering was taken into account, we can estimate that for an evacuated flightpath before and after the sample, Ar atmosphere at the sample position, two 2 mm thick sapphire windows before and after the sample, and a total of 4 mm aluminum in the beam, the losses for 25 Å neutrons to be 25%-35%.(The total transmission is derived as the product of each transmission FIG. 6.A reflectivity simulation from a free D 2 O surface using the grid absorber for 8 distinct incident beams; bars indicate the measurement errors after 1 s exposure.The vertical scale on the right indicates the intensity incident on a 20 × 50 mm 2 sample for each beam normalized by the typical reflectivity drop of q −4 z .The integrated intensity for the 6 lower beams is less than 5% of the 3.0 • beam and thus does not significantly contribute to the measurement background. component which itself is derived from the attenuation coefficients from the noted publications: 30 cm Ar 16 T Ar = 0.93, 4 mm sapphire 17 T Sa = 0.99, and 4 mm Al 18 T Al = 0.8.)On the scale presented in Fig. 6, this reduction will not be relevant.
In addition to the simulated reflectivity, Fig. 6 also shows the incident neutron flux for each beam normalized by the typical drop of reflectivity of q −4 z to illustrate the statistics that can be expected in the different q z regions.As the first 5 beams have higher normalized intensity than any point at higher q z , the reflectivity below 0.015 Å ☞1 in this mode will always have sufficient statistics.For beams 6-8, the intensity is almost flat after the normalization, varying by about a factor of 3 for q z up to 0.2 Å ☞1 .Most experiments that would require this and an extended q z range would therefore have sufficient statistics everywhere, if the highest bins fulfill this criterion.For increased significance in the intermediate range, beams 6 and 7 could be set to lower angular resolution to closer match the high q z measurement.
In the case of GISANS and to increase the signal to background ratio for specular and OS, the hyperbolic mirror will be employed to shift the focus of the beam from the sample position onto the detector.With the reflector starting at 3 m and ending 0.5 m before and detector at 3 m after the sample, this leads to a maximum usable sample size of approximately 20 mm.In this configuration, the coma aberration from the elliptic neutron guide is not fully compensated by the hyperbolic reflector, leading to an angle dependent size of the beam if the VS is a conventional slit.However, using a position offset for the left and right absorber of the VS allows this effect to be mostly compensated as the slit apparently has different widths for neutrons with different directions.Figure 7 shows the beam intensity on the detector for three (average apparent) sizes of the VS slit.While the size of the source is magnified by about a factor of 4, the intensity profile for all settings is almost flat with sharp boundaries.Thus the setup allows a free trade-off between resolution and intensity in a FIG. 7. Simulated direct beam on the detector for three sizes of the horizontal VS with about 4 times magnification from the guide.The VS left and right absorbers are offset in the beam direction to compensate aberration, and thus the resulting resolution functions are well defined.Beam resolutions down to ≈1 mm can be achieved for a resolution at 6 Å of ∆q y = 3.5 × 10 ☞4 Å ☞1 covering up to q y = 0.1 Å ☞1 .wide range between 1 mm and 10 mm while keeping the full sample area illuminated.
The magnification effect of refocusing is, in first order, given by the ratio between the reflector-to-sample and reflector-to-detector distance multiplied by the magnification due to the elliptic guide VS-to-guide to guide-to-sample ratio.In the HEKATE case, the ellipse has a magnification between 1:3 and 3:1 for beams reflected at the beginning or end of the guide, respectively.This leads to a total magnification of 1:6 and 3:7.5, which was compensated by the asymmetric VS to about 1:4 for all divergence angles.For samples smaller than the beam footprint, the magnification can be reduced by using a smaller sample to detector distance, which not only allows an increase of intensity for the same q resolution but also increases the measured q range.(That is, a sample of 5 mm width could be measured at 1 m detector distance; the necessary reduction in VS size of about 50% would be compensated by a 4 times increase from the reduced beam size while, at the same time, tripling the q range.)

IV. COMPARISON TO EXISTING INSTRUMENTS
The simulations carried out with HEKATE can directly be compared to the performance of existing instruments of the reflectometry and classes.As a benchmark, we have chosen the TOF instrument FIGARO on the ILL reactor source in Grenoble and the EQ-SANS instrument at the Spallation Neutron Source (SNS) in Oak Ridge, TN, as these facilities are at the forefront of today's neutron science.
Comparing the GISANS mode flux shown in Fig. 5 with the data from EQ-SANS (Fig. 9 in Ref. 19) that was measured with a comparable resolution (40 × 40 mm 2 aperture with 4 m collimation to 30 × 30 mm 2 beam size at 3 m detector), one finds a ≈50 fold increase.At the same time, the EQ-SANS bandwidth is only 3 Å at the highest resolution of 2 × 10 ☞3 Å ☞1 in contrast to the 10.5 Å band and about 10 times better achievable resolution of HEKATE.FIGARO is a reflectometer optimized for measurements on free liquid surfaces.Measurements performed on free D 2 O that use relaxed resolution for fast acquisition have been published in Ref. 20 (Fig. 10).The q z -range covered in this measurement was 0.005 Å ☞1 to 0.25 Å ☞1 , and it took 150 s to record.For HEKATE, the reflectivity was simulated with the grid method as shown in Fig. 6 using only 1 s exposure and achieving a better statistics, larger q z -range and below 0.05 Å ☞1 even better resolution.In addition to these direct comparisons, we have used the neutron flux at the sample position for different wavelength resolutions and normalized this value to an angular resolution that would achieve a minimum q y of 0.01 Å ☞1 , assuming a quadratic dependence of the intensity and the divergence (horizontal times vertical).Table I summarizes the comparison of the flux simulated for HEKATE with values published in Refs.21-25.Further the ESS instruments already in the design phase have produced simulated results to benchmark the instrument performance.We only consider the reflectometer FREIA and the SANS instrument SKADI, 26 as Estia does not allow measurements from free liquids and LoKI has much relaxed resolution.The FREIA reflectometer allows measurements with 0.2 • -4.0 • incident angle and has a flux maximum at ≈3.5 Å, leading to similar maximum q z as HEKATE.The moderatorto-detector distance is ≥25 m, and thus the natural resolution of both instruments is similar, as well.Comparing Fig. 5 with Fig. 12(b) in Ref. 27, the uncollimated flux of FREIA at the sample position, one finds that the peak at 4 Å is 5 × 10 7 s ☞1 cm ☞2 Å ☞1 , a factor of 10 smaller than that in the HEKATE case for guide b as a result of lower brilliance transfer and horizontal divergence.The HEKATE guides cannot serve sample widths above 20 mm but will still allow most reflectometry and OS experiments.To compare our concept to the SKADI SANS instrument, we have to translate the available intensity information to a GISANS experiment on a 10 × 50 mm 2 sample with 20 m collimation and incident beam around the critical edge of total reflection.We find a neutron flux on the sample of 3 × 10 5 s ☞1 which compares to the HEKATE with the same q-resolution yielding 1.4 × 10 5 s ☞1 or 2.0 × 10 5 s ☞1 in the duplex mode, considering the larger wavelength used and that the guide has not been systematically optimized, a satisfactory result.Additionally, the necessary focus size was 6 mm so that the HEKATE instrument could reach up to 6 times better GISANS resolution, if desired.For the same resolution, the conventional pinhole geometry would require either ≈60 m collimation length or significantly reduced beam size.For more information about the instrument comparison, see the supplementary material.

V. CONCLUSION
In summary, we have presented a novel neutron guide and wavelength band definition concept for GISANS and adapted it to the ESS requirements.This led to the presented instrument study HEKATE.Separation of the neutron guide into two distinct pathways allows us to efficiently transport short wavelengths to hit the sample under high incident angle while, at the same time, supplying a large divergence beam with longer wavelengths to lower incidence angles.The combined usage of focusing reflectometry with three separated wavelength frames that are distinguished on the detector reaches unprecedented q z range coverage from 10 ☞3 Å ☞1 to 0.3 Å ☞1 , even from free liquid surfaces.For dedicated GISANS experiments, the focusing optics reach high q resolution in the sample plane and the duplex mode with free wavelength band distribution allows high flexibility to concentrate on the sample specific details.Finally, the multiple line of sight avoidance, small monolith opening, and early beam size reduction at the virtual source will reduce the background from high energy particles to a minimum.

SUPPLEMENTARY MATERIAL
See supplementary material for a discussion of the necessary resolution in GISANS with application examples, simulation of focusing reflectometry, and details about the comparison to ESS instruments.

FIG. 3 .
FIG.3.TOF diagram for the wide q range reflectometry mode.The three bands selected by the different openings in a single chopper reach distinct positions on the detector and can thus be separated in the data analysis.Colors for the frames are matched to the ones used in Fig.1, and the horizontal bars at 7 m denote the closing times of the different chopper regions.

fully within the normal ESS monolith beam extraction
• ,

TABLE I .
Comparison of total flux at the sample position in units of neutrons/(s × cm 2 ) for different wavelength resolution extrapolated to an instrument setup with a minimum q y -resolution of 0.01 Å ☞1 .