Design and performance of the multiplexing spectrometer CAMEA
Jakob Lass, Dieter Graf, Felix Groitl, Christian Kägi, Raphael Müller, Roman Bürge, Marcel Schild, Manuel S. Lehmann, Alex Bollhalder, Peter Keller, Marek Bartkowiak, Uwe Filges, Frank Herzog, Urs Greuter, Gerd Theidel, Luc Testa, Virgile Favre, Henrik M. Rønnow, Christof Niedermayer
DDesign and performance of the multiplexing spectrometer CAMEA
Jakob Lass,
1, 2
Dieter Graf, Felix Groitl,
1, 4
Christian Kägi, Raphael Müller, Roman Bürge, Marcel Schild, Manuel S. Lehmann, Alex Bollhalder, Peter Keller, Marek Bartkowiak, Uwe Filges, Frank Herzog, Urs Greuter, Gerd Theidel, Luc Testa, Virgile Favre, Henrik M. Rønnow, and Christof Niedermayer Laboratory for Neutron Scattering and Imaging,Paul Scherrer Institut, CH-5232 Villigen, Switzerland Nanoscience Center, Niels Bohr Institute, University of Copenhagen, DK-2100 Copenhagen Ø , Denmark Laboratory for Neutron and Muon Instrumentation,Paul Scherrer Institute, CH-5232 Villigen, Switzerland Laboratory for Quantum Magnetism, Ecole Polytechnique Fédérale de Lausanne (EPFL), CH-1015 Lausanne, Switzerland Laboratory for Particle Physics, Paul Scherrer Institute, CH-5232 Villigen, Switzerland (Dated: July 30, 2020)The cold neutron multiplexing secondary spectrometer CAMEA (Continuous Angle Multiple En-ergy Analysis) was commissioned at the Swiss spallation neutron source SINQ at the Paul ScherrerInstitut at the end of 2018. The spectrometer is optimised for an efficient data collection in thehorizontal scattering plane, allowing for detailed and rapid mapping of excitations under extremeconditions. The novel design consists of consecutive, upward scattering analyzer arcs underneathan array of position sensitive detectors mounted inside a low permeability stainless-steel vacuumvessel. The construction of the world’s first continuous angle multiple energy analysis instrumentrequired novel solutions to many technical challenges, including analyzer mounting, vacuum connec-tors, and instrument movement. These were solved by extensive prototype experiments and in-housedevelopments. Here we present a technical overview of the spectrometer describing in detail the en-gineering solutions and present our first experimental data taken during the commissioning. Ourresults demonstrate the tremendous gains in data collection rate for this novel type of spectrometerdesign.
Keywords: Neutron Scattering Instrument, Spectroscopy, Massive Multiplexing Instrument, Inelastic Neu-tron Scattering
I. INTRODUCTION
Triple-axis spectrometers are workhorses of modernneutron scattering experiments. They allow mappingof both the static correlations and the elemental excita-tions in condensed matter materials, which provide fun-damental insight into the microscopic interactions of themeasured systems. Neutron scattering experiments areusually flux limited, which leads to continuous efforts toimprove the experimental setup. These consists of in-creasing the number of neutrons created at the source,improving the number of neutrons that reach the sam-ple, and optimizing the detection efficiency of scatteredneutrons.In this publication we report on the design andcommissioning of the new multiplexing spectrometerCAMEA[1], replacing the secondary spectrometer partof RITA-II at SINQ. CAMEA possesses 104 position sen-sitive detectors, which are collecting scattered neutronsfrom 600 analyzer crystals enabling a coverage of a largepart of reciprocal space within the horizontal scatter-ing plane. This setup is ideal for experiments that re-quire extreme environments, such as high magnetic fieldand pressure. The concomitant increase in data requir-ing reconsideration of the instrument control system, thedata processing procedure and data handling. These ob-stacles were overcome with the new software packageMJOLNIR[2]. Finally, we report first experimental re-sults obtained with the new spectrometer, demonstrating the enormous capacity of the CAMEA concept. Beingthe first instrument to utilise the prismatic analyzer con-cept [3] and due to the similarities with the upcomingBIFROST spectrometer, currently under development,at the European Spallation Source[4], CAMEA will pavethe way for the next generation of massively multiplexingsecondary triple axis spectrometers.It is of importance to state that the introduction ofmassive multiplexing mapping instruments does not re-move the raison d’être for standard triple axis instru-ments, as these still allow for a more detailed parameterstudies where overview maps of Q and ∆ E are less im-portant. II. IMPLEMENTATION
CAMEA has replaced the RITA-2 multiplexing instru-ment located at the neutron guide port RNR13 at SINQ.During 2019 and 2020 SINQ undergoes a major upgradeprogram of its neutron guide system. The primary spec-trometer of CAMEA will consist of new guides with con-verging elliptical sections reaching m values up to a factorof 4, focusing the neutrons on a virtual source placed 1.6m upstream of a double-focusing monochromator. Thisupgrade will result in a flux gain of up to a factor of5 at the relevant energies, while undesired neutrons aresuppressed. The basic concept of the secondary spec-trometer has been described in detail by Groitl et al. [1]. a r X i v : . [ phy s i c s . i n s - d e t ] J u l In this publication we will give an account of key fea-tures and focus on the technical implementation of theCAMEA concept.
A. Cross-sectional overview of the secondaryspectrometer
Fig. 1 shows a cross section view into the stainless steelvacuum vessel of the secondary spectrometer. It consistsof 8 identical analyzer modules that cover 61 degrees inscattering angle. Each module consists of 8 focusing an-alyzers angled to scatter neutrons at fixed final energiesranging from 3.2 meV to 5 meV, see table I. A beryllium(Be) filter with a build-in radial collimation of 1’ is in-stalled before the analyzer array[5]. This unit is cooledto temperatures below 80 K by a Gifford McMahon cry-ocooler. At this temperature the transmission of the Beis close to 100 % for neutron energies below 5.2 meV.The low temperature ensures suppression of Be phononsallowing a cleaner filtering. The upwards scattered neu-trons are detected by a radial layout of 104 position sen-sitive He detectors tubes, all 1/2 inch in diameter. Dueto their radial placement they are arranged in a stack oftwo staggered layers. Cross talk is prevented by Boral-can shielding separating individual wedges and analyzerswithin the wedges. All components are mounted withina shielded non-magnetic stainless steel vacuum tank. Inthe following sections we give an an in-depth descriptionof the main components.
FIG. 1. Sectional view of the components inside the vacuumtank, showing the Be filter on the left of the 8 analyzers andthe beam stop to the right. Cross talk shielding is placedbetween the analyzers and the detectors. The detector elec-tronics is mounted outside of the vacuum on top of the tank.
B. Vacuum tank
Several recursions were taken to minimize sources ofbackground. As such, the entire CAMEA unit has beenenclosed in a vacuum tank that minimizes air scattering.The vacuum tank is made out of stainless steel, forwhich individual pieces were carefully selected with re-gard to their magnetic permeability. Only pieces with µ < 1.05 10 − H/m were chosen in the manufacturingprocess. Specialised welding material was used to ensurethat the tank retained its non-magnetic state. High mag-netic field tests using an external vertical magnetic fieldof up to 13.5 T were performed during commissioningconfirming similar conditions as for the replaced RITA-2instrument. No changes are expected up to 15 T, whichis currently the maximum field available at SINQ. Thetank wa evacuated at room temperature to a pressure of − mbar. At this pressure ,the cryocoller was switchedon to cool the Be filter to its base temperature of 65 K.At this stage, a pressure of × − mbar was reachedinside the vacuum tank. FIG. 2.
Left : The stainless steal vacuum tank on top ofthe support structure made of Aluminium before the 200 mmthick polyethylene blocks were added.
Right : The CAMEAinstrument in operation using the MA15 cryomagnet at a ver-tical field of 13.5 T.
C. Shielding
The Pb target of SINQ produces fast neutrons withenergies up to the proton beam energy of 570 MeV. De-spite shielding around the target structure, some of thesemay enter the detector tank, if no additional shielding isused. We, this added 200 mm thick borated polyethy-lene blocks, with a contents of 5% BN, to the outsideof the tank. The blocks moderate fast neutrons to thethermal range where the boron content efficiently cap-tures the neutrons. Neutrons scattered by the sample,enter the tank through an Aluminium window. Depend-ing on their energies they will be scattered upwards bythe different analyzer modules. CAMEA has an openanalyzer and detector geometry. Thus, cross-talk be-tween the different analyzers and wedges which needsto be suppressed. This was achieved by inserting a 3mm thick honeycomb-like structure made of Aluminiummixed with 20% B C (Boralcan) between all wedges andanalyzers. This guarantees that only neutrons scatteredat a particular analyzer segments reach the designatedpart of the detectors. In addition, the whole tank is cov-ered with boralcan. Our combined shielding provisionsresult in a low background of about 0.3 counts per minuteper detector when the main shutter is closed. This num-ber, thus, includes dark currents and background fromthe environment. During operation, after the neutronsenter through the Aluminium windows they pass throughour Beryllium filter, which acts to remove neutrons withenergies above 5.2 meV. These scattered neutrons are ab-sorbed by Cadmium in the filter, generating gamma radi-ation. To suppress these, a Pb shielding has been addedon top of the filter. Similarly, neutrons not scattered bythe analyzer array end up in the beam stops located be-hind all of the analyzers. Here, gamma radiation is alsogenerated due to neutron absorption and a Pb shieldinghas been put in place.
FIG. 3.
Left : Cross-talk shielding between analyzers andwedges prevent contamination of signals from the differentenergy and angular channels.
Right : Cross-talk shieldinginside the vacuum tank.
D. Air bearing
Conventional triple-axis instruments are moved oftenduring experiments. This is necessary due to changeof incident energy affecting the scattering geometry. Inmost cases this is achieved via pressurised air-pads thatallow instrument to glide over a polished marble floor.CAMEA is based on the same principle and an efficientlymethod to reposition the vacuum tank is needed. Theenormous weight of CAMEA requires the development ofa new suspension system, distributing the weight evenlyacross the large tank, while ensuring a smooth movementwith high precision. We used an assembly of air-padsfrom AeroLas as shown in Fig. 4. Vibrations were pre-vented by the use of large cushion elements made out ofsteel wool inserted between the air-pad and its mounting.
FIG. 4.
Left : Close up of air pad with steel wool used as adamping element.
Right : 3D rendering of air bearing solu-tion.
E. Be-filter, cooling and performance
The Be filter between sample and analyzer is a nec-essary requirement to reduce parasitic scattering as wellas higher order scattering in sample and monochroma-tor. Since the scattered neutron should be collimated di-rectly after the sample and due to spacial restrictions, acombined radial collimator and Be filter [5] was designedand manufactured. The optimal design consists of plac-ing thin Be slices mounted between 0.3 mm thick glasslamellas coated with B. The spacing of the lamellas waschosen to provide a collimation of 1 degree. The filter-collimator combination was mounted within the vacuumvessel and dry cooled with a cryo cooler providing a cool-ing power of 175 W at 77 K. We reached a cooling rateof about 5 K per hour, resulting in a cooling time of lessthan 2 days. After reaching a base temperature of 65 Kthe temperature slowly increased at a rate of about 1 K/day. The loss in cooling power is believed to be caused byresidual gas on the large surface of the collimator, result-ing in a reduced emmissivity. This problem can be solvedby installing a sorbitol pump next to the cold head.
FIG. 5.
Left : The combined Be filter collimator unit consist-ing of 8 different segments.
Right : Cold copper head of theBe filter.
F. Analyzer Design
The analyzer arrangement consists of 8 identical mod-ules with a total angular spread of 61 degrees. Tech-nical details of the design are described by Groitl et.al. [1]. The analyzer arcs are composed of 600 highlyoriented pyrolytic graphite (HOPG) crystals (PanasonicPGCX07SP). The quality of all crystals was examinedby neutron diffraction on the thermal time-of-flight neu-tron diffractiometer POLDI. Among different mount-ing possibilities[6], we chose a purely mechanical optionbased on clips and stoppers simply holding the crystals inplace. These are made out of Anticorodal AC110, a stifferaluminium alloy, that can be cut precisely. This allowsmanufacturing of holders and clips with high precisionas to circumvent the cumbersome and error prone pro-cess of crystal aligning either through laser reflection orneutron diffraction. The 3D design of clip solution usedat CAMEA is shown in Fig. 6 together with a pictureof a showcase example of Cu mounted like the graphitepiece. The height of the clip and stopper were chosen toexert sufficient force onto the HOPG crystals and the Siwafer to ensure that they do not slide during tank move-ment. Moreover, the force needed to be low enough, suchthat the crystals were not bend, changing their proper-ties. Two clips of 2 mm in width were combined withthe clamp at the end of the clip further ensuring that theHOPG does not tilt or move during experiments.
FIG. 6.
Left : 3D rendering of the analyzer mounting so-lution, consisting of clips and stoppers to attach the HOPGcrystals to Si wafers.
Middle : Clip and holder design usedto secure the HOPG crystals onto the Si wafers and these tothe aluminium frame.
Right : Showcase of Cu mounted likethe HOPG crystals of the analyzers.
G. Detectors and Plugs
The detector system of CAMEA consists of 104 po-sition sensitive half-inch detector tubes (Reuter Stokes)with a length of 1 m. Each tube is filled to a total pres-sure of 9.27 atm which includes 7.1 atm of He. A radialarrangement in two planes was used resembling a ’W’pattern as can be seen in Fig. 7. The configuration isdesigned to cover as much area above the analyzers aspossible while minimising gaps between the tubes. Thischoice results in detector tubes from the two layers over-lapping towards the sample position, which affects thesensitivity of neutron detection for the upper layer. Thisis to be taken into account in the normalization proce-dure described in section III A. Despite the optimisation,gaps between detector wedges are unavoidable leading toan angular coverage between 50% and 70% for low andhigh final energies, respectively[1].CAMEA is designed in a modular fashion by 8 almostidentical wedges, each consisting of 13 detector tubes lo-cated above 8 analyzer arcs. The ’W’ configuration is realised by alternating 7 and 6 detector tubes in the up-per and lower layer, respectively, see Fig. 7.
FIG. 7.
Left : Radially arranged detectors tubes.
Right : In-plane detector arrangement of two nearby segments, showingthe ’W’ configuration.
The radial arrangement of the detector tubes requiresa very compact design of transition pieces and plugs lessthan 0.5 inches and could thus not be used for our de-tector design. Our newly designed connectors are coaxialand are based on the standard Lemo plug, which can beoperated up to a voltage of 2 kV. Further, these do notsuffer from leakage currents for the measurement.The spacial sensitivity of the detector tubes is achievedby recording the relative charge distribution reachingeach end of the tube. Connectors and coaxial cablelengths are identical on each side end of the detector,which guarantees a homogeneous impedance which iseasier to handle by the electronics. Coaxial conductorsalso posses substantial shielding and an insulation jacket,which offer ideal conditions for preventing cross-talk be-tween adjacent conductors. Kapton (orange colour inFig. 7) was chosen as insulating material, since it doesnot degrade noticeably in vacuum.With the chosen design, the detectors are fully func-tional at the operation pressure of 10 − mbars. However,if the pressure increases above 10 − mbar, the detectorsHV supply needs to be shut down due to the Paschenregime[7]. H. Detector electronics
Neutron events along the tube axis are measured usinga resistive charge division[8] readout concept. Each tubeis read out at both ends, requiring 208 analogue acqui-sition channels for the CAMEA instrument. A new gen-eration of readout electronics was designed to achieve ahigh integration density. The key components for this de-velopment are highly integrated multi-channel Analogue-to-Digital Converters (ADCs), serialised high-speed datainterfaces, and Field Programmable Gate Arrays (FP-GAs).A simplified block diagram of the readout electronics isshown in Fig. 8. The electric charge collected at the de-tector tube ends is amplified by means of transimpedanceamplifiers, one for each end of the tube. The amplifiedsignals are low-pass filtered and driven into an ADC. TheADC operates at a sampling rate of 50 Mega-SamplesPer Second (MSPS) per channel and with a resolution of14 bits. The digitised sample data flows into an FPGA,which extracts the pulse event information from the con-tinuous ADC data stream. A data packet is generated foreach pulse event, containing a source identifier, a times-tamp, the actual position information, and the chargecontent of the event, i.e. the summed sample valuesabove the trigger threshold. Finally, the event packetsare collected into User Datagram Protocol (UDP) framesand sent to the generic computing infrastructure via anoptical Small Form-factor Pluggable transceiver (SFP)module.
FIG. 8. Simplified readout electronics block diagram. In theCAMEA setup each ADC chip contains 16 parallel conver-sion channels, with two ADC chips connected to one FPGA.Hence, each system provides 32 analogue acquisition channels.
Commanding and updating of the readout system takeplace through the same SFP module, except that theTransmission Control Protocol (TCP) is used to ensurehigher reliability. Many system parameters, e.g. thesignal-trigger threshold and the gamma rejection level,are mapped to programmable registers in the FPGA. Inaddition, a test pulser is included for debug purposes,which was helpful in the early stages of instrument com-missioning. It allowed us to provide a well-defined inputto the data collection, histogramming and data displayalgorithms before the startup of the full instrument. Afurther debug feature is the extraction of raw ADC wave-form data for neutron events, enabeling checks of stim-ulus and response verification of the trigger, peak detec-tion, and position calculation functions.The actual hardware, denoted the CAMEA Front-endBox, is shown in Fig. 9. We chose a modular design inwhich transimpedance amplifiers are self-contained plug-in modules with connector interfaces to the ADC PrintedCircuit Board (PCB). This design further allow for reuseof parts of the system for future instruments. The ADCPCB carries 16 amplifier modules (8 on each side) pro-viding a similar amount of analogue input channels. Themodule also contains a high voltage bias injection with anappropriate conductor clearance to prevent arcing. Mostof the digital functionality is contained on the data con-centrator board, namely the FPGA, the clock source, theconfiguration logic, and the digital interfaces. High-speedserialised data lanes connect the ADCs to the FPGA.The continuous data generation rate of the two ADCsamounts to 22.8 Gbit/s. We chose a Gigabit Ethernetconnection for the outgoing data link. It is, however,noted that up to 24 Gbit/s could be implemented using a Quad SFP (QSFP) module.
FIG. 9.
Top : ADC board with plug-in amplifier modules.
Middle : Data concentrator board.
Bottom : AssembledCAMEA Front-end Box with sidewalls and removed top cover.
Prior to the installation of the electronics boxes atCAMEA, the design was tested at MORPHEUS at SINQ.MORPHEUS allows a narrow beam of neutrons to be di-rected onto well-defined tube positions which is not pos-sible in the final CAMEA instrument assembly. Here, thedetector tubes are always out of the direct beam. Fig. 10shows the results of the test campaign. The positionalresolution expressed in terms of the Full-Width at HalfMaximum (FWHM) of the pulse position histogram isbelow 5 mm in the central tube area of the tube, and be-low 10 mm towards the endpoints. A degradation of theresolution towards the two ends of the tube is expectedfor a charge division readout system as most of the chargeflows to the near end whereas a close-to-threshold signalis detected at the far end with a correspondingly lowSignal-to-Noise Ratio (SNR). These results allow estima-tion of a lower limit for individual pixel size along thedetector tube. We chose a minimal pixel size of about ∼ FIG. 10.
Top : Position measurement histogram of a spotbeam scan. The translation from leftmost to rightmost posi-tion is 825 mm, with a step size of 55 mm.
Bottom : Pulseheight spectra of the 13 detector tubes attached to one front-end box.
III. DATA TREATMENT AND REDUCTION
In this section we discuss the treatment of the raw dataand the process of converting the detector counts fromthe 1024 different pixels of each of the 104 position sen-sitive detector tubes into scattering intensity as function of momentum transfer Q and energy transfer ∆ E . Theprocess requires determination of the scattering angle A (the angle between incident and scattered neutron beam)and the energy transfer (the energy loss or gained duringthe scattering process). In the latter case, we note thatthe incoming energy is defined by the monochromatorsetting, while the final energy at a given pixel positionneeds to be determined from the detector-analyzer setup.A sensible data treatment also requires construction ofa normalization matrix, correcting for inhomogeneities indetection efficiency among all 1024 x 104 pixels. A. Normalization energy scans with Vanadium
One efficient way to normalize all analyzer-detectorpairs is to measure a strong incoherent scatterer suchas Vanadium. Measuring the scattering intensities as afunction of incoming energy allows determining the en-ergy dependence of the detector pixels and their rela-tive sensitivity. The upwards scattering analyzer arcsare designed to measure 8 different final energies rangingfrom 3.2 to 5.0 meV[1]. Because the corresponding eightanlayzers are geometrically fixed, incoherently scatteredneutrons of a given energy will be confined to 8 well-defined pixel areas on each individual tube detector (seeFig. 11). P i x e l n o . Summed P i x e l n o . Summed
FIG. 11. Comparison between
Top : Vanadium normalisationscan. and
Bottom : simulated McStas model.
In a first step, the detector area corresponding to the8 final energies needs to be determined for each indi-vidual tube.This can be achieved scanning the incomingenergy from 2.9 meV and 5.5 meV. Considering the in-cident wavelength spread, reproducibility of monochro-mator angle and the amount of data points required onthe detector, an optimum of 250 to 500 scan points wasfound (cf. Figs.12 and 13).The resulting data is a 3D scattering intensity matrixwith the axis being incoming energy (i.e. scan step), de-tector number, and location along detector. Summingover incoming energy, one of the normalization scans isshown in Fig. 11. The neutron scattering intensity perpixel is shown in a 2D plot of pixel versus detector num-ber. We also show the calculated intensity determinedby a McStas[9] simulation. Note the very satisfactoryagreement between experimental and simulated data con-firms the geometrical accuracy of our analyser-detectorsetup. The fixed design of the secondary spectrometeralso promises a long term stability of the intensity distri-bution on the detectors.Depending on the integration limits of the incomingenergy, different parts of the detector tubes have highintensity.Fig. 12 shows that each detector tube has 8 high in-tensity areas, separated 7 low intensity regions. Theycorrespond to neutrons scattered from the eight focusinganalysers underneath the tubes and the regions betweenthe analysers, respectively. We denote the prior as de-tector segment. The excellent separation between the in-dividual segments confirms the quality of the cross-talkshielding. The intensity distribution along each detectortube is be fitted by eight. As a representative examplewe show the result to such a fit to the data in detector41 in Fig. 12.The red points correspond to an acceptance width of ± σ . The low intensity regions are displayed in black.A 3 σ acceptance region results in a suitable compromisebetween maximal signal and background. During dataprocessing the low intensity areas are masked to reducenoise and background not originating from the sample. Pixel Number N e u t r o n C o un t FIG. 12. Intensity in detector tube 41 summed across incom-ing energy. Red dots denote pixels within an acceptance areaof ± σ . Finally, the individual pixels are assigned to final ener-gies. The normalization scan has a dimensionality of 501 (scan steps) ×
104 (detectors) × × ×
8. For each segment the intensity isnow known with respect to the incoming energy (Fig. 13for the eight segments of detector 41). Each peak is fit-ted by a Gaussian distribution extracting the nominalenergy, resolution, sensitivity, and background level. E f [meV] B i nn e d N e u t r o n C o un t FIG. 13. Binned neutron count in the active detector areas asfunction of incoming energy for detector 41 in the Vanadiumscan.
We note that the derived nominal energies and widthsare stored in the data files together with the background,amplitude, and pixel edges. This facilitates the datatreatment for users of the instrument. It also ensuresthat the required instrument parameters are included inevery single data file, such that they can be analysed in-dependently and without the need of a specific Vanadiumnormalization scan.
B. Scattering angle
The calibration of the momentum transfer requires thescattering angle for every pixel for all the detector tubeswith respect to the incoming beam direction. The tubeshave both a relative scattering angle among themselves,denoted A , and an absolute angle depending on the de-tector tank rotation, θ .In strong contrast to standard TAS, the scattering an-gle at CAMEA needs to be determined on a pixle bypixle basis. This is due to the out-of-plane scatteringcombined with flat analyzers and detector tubes placedradially above these. A determination can be achievedeither experimentally through a set of measurements orby a calculation based on the geometry of the analyser-detector system. The former method requires 8 angularscans for each of the 8 final energies on a sample withlarge lattice parameters. The latter relies on an accuratedescription of the instrument.During the commissioning phase we relied on calcula-tions, but checked their validity with two energy scans atthe highest and lowest final energy. We used a high qual-ity Pr Hf O single crystal for which we found excellentagreement with our calculations, less than 1 deg. IV. INSTRUMENT PERFORMANCEA. Energy and A4 resolution
A summary of calculated and observed instrument pa-rameters is shown in table I.A comparison to the measured parameters, given intable I, to the design specifications for the secondaryspectrometer alone is tabulated in table III. Here a dis-crepancy is clear. While the final energies are withinthe expected range, their resolutions are on average 40%worse than our calculations for the back-end. This dif-ference has been further investigated by a McStas sim-ulation with a 1 cm high and 1 cm diameter cylindricalVanadium sample in the current experimental setup. Theenergy resolution was found to agree within 4 % and itis concluded that the primary spectrometer is the mainsource of energy broadening.The noted A4 checks of the secondary spectrometerreveal average FWHM of 0.5451 deg for 5 meV. Thiscorresponds to an uncertainty of 0.014 reciprocal latticeunits (rlu) using a cubic crystal with a = 2 π . This iswell within the performance requirements of triple-axisinstruments. B. Prismatic concept
The CAMEA concept allows a coverage of large partsof the reciprocal space positions in a single acquisitionscan, but the position sensitivity of the detector tubes arealso useful to improve the energy resolution. The pris-matic concept[3] exploits the distance collimation of thesecondary spectrometer. If cases where both the sampleand the detector are small, the energy distribution on thedetector is narrow, and the detection rate is low. By re-laxing the mosaicity of the analyser crystals the analyserenergy distribution is broadened but the detector energyresolution remains the same as it is governed by distancecollimation. This ultimately yields an increase in neutroncount at minimal cost to the energy resolution.The prismatic concept fully comes to live at CAMEAdue to the position sensitive detectors used across theback-end. This allows a subdivision of the active areasegments of the 8 energies into smaller segments. Thatis, the otherwise discarded energy distribution within thesegments is used to improve the energy resolution. As anexample choosing 5 sub-pixels is equivalent to placing 5smaller detectors closely along the detector tube yielding5 energies instead of one. This impacts the intensity asthe total neutron count is of course unchanged. This sub-division modifies the normalization procedure in the stepwhere the nominal energies are found and subsequently their normalizations. Here, the intensity of every energysegment is split into n subpixels, which distributes theneutron counts over 1024/(8 n ) pixels. This is apprecia-bly more sensitive compared to summing the intensityover 8 segments only. Thus, users can in principle choosethe energies measured without compromising the totalintensity.It is here noted that a too aggressive subdivision canresult in oversampling, for which systematic errors in theinstrument resolution gain in importance. In this regimethe energy resolution cannot be improved further, butan increase in n yield low count bins that are artificiallyseparated. Our measurements of the detector character-istics (see section II G) it was found that the positionsensitivity of the detector tubes is around 5 mm on 900mm. Considering the 1024 pixels of the raw signal, thiscorresponds to a minimal sub-pixel size of 5 pixels.In Fig. 14 we show the prismatic concept by plottingthe 5 t h energy segment of detector 41 with n = 1, 3 and5. The sub-pixels are simply found by linearly splittingthe active area of the segment. The resulting energy-dependent signal for these three binnings is shown to theright in Fig. 14 and the nominal energy and width istabulated in Table. II.
500 520 540 560 580 600 620
Pixel Number N e u t r o n C o un t InactiveBin 0 E f [meV] N e u t r o n C o un t
500 520 540 560 580 600 620
Pixel Number N e u t r o n C o un t InactiveBin 0Bin 1Bin 2 E f [meV] N e u t r o n C o un t
500 520 540 560 580 600 620
Pixel Number N e u t r o n C o un t InactiveBin 0Bin 1Bin 2Bin 3Bin 4 E f [meV] N e u t r o n C o un t FIG. 14.
Left : Pixels used when binning into 1, 3, and 5 binsper segment for detector 41 and segment 5.
Right : Measuredand fitted energy dependency using the same bins.
Channel 1 2 3 4 5 6 7 8 E f Calculated [meV] 3.200 3.374 3.568 3.787 4.033 4.313 4.629 4.993 E f Measured [meV] 3.179 3.361 3.552 3.764 4.010 4.290 4.605 4.963 ∆ E f FWHM Measured [ µ eV] 147 159 171 186 203 227 253 291 k f Measured [1/Å] 1.239 1.273 1.309 1.348 1.391 1.439 1.491 1.548Take off angle θ [degree] 98.2 94.7 91.3 88.0 84.6 81.2 77.8 74.5 V Resolution
Measured [1/Å ] 1.645 1.904 2.193 2.534 2.959 3.475 4.104 4.877Normalization factor Measured 1.00 1.08 1.19 1.26 1.28 1.27 1.27 1.24TABLE I. Analyzer parameters for CAMEA as measured during the hot commissioning phase. k f , θ and resolution volumewere calculated from E f . Normalization factor is relative to channel 1.Segment 1E f [meV] 4.006FWHM [ µ eV] 202Segment 1 2 3E f [meV] 3.924 4.006 4.095FWHM [ µ eV] 133 145 147Segment 1 2 3 4 5E f [meV] 3.893 3.949 4.009 4.064 4.134FWHM [ µ eV] 126 128 134 137 146TABLE II. Final energy and FWHM for segment 5 for detec-tor 41 in commission. A re-binning of the data using sub-pixels significantlyimpacts the energy resolution, see Table. II. In Fig. 15 weshow the energy dependence of the energy resolution fora single detector tube, using the FWHM of the Vanadiumnormalization scan.The data are over-plotted with a McStas simulation,considering only the secondary spectrometer. It has beenperformed using a cubic 1 cm source at the sample po-sition with a divergence of 2 degrees. The simulationreveals the best possible resolution independent of theguide and monochromator. The results justify the needfor an upgrade of the primary spectrometer to match theresolution of the secondary spectrometer. This upgradeis being performed in 2019/2020. It consists of a newguide with a scalable virtual source and a double focus-ing monochromator. E f [meV] F W H M [ μ e V ] File 115 - Bin 1File 115 - Bin 3File 115 - Bin 5Simulation - Bin 1Simulation - Bin 3Simulation - Bin 5
FIG. 15. Mean FWHM for final energy at elastic line for n equal 1, 3, and 5 over-plotted with McStas simulations con-sidering only the secondary spectrometer. The linear fits areguides to the eye. C. Spin waves in MnF2
The first commissioning phase allowed testing CAMEAunder various setup, covering different scientific topics.Most results will be communicated in separate publica-tions [10][11]. Here, we focus on an experiment using aMnF single crystal of 6.2 g. The the spin waves for thisspin 5/2 compound have been extensively studied[12],which makes this crystal an ideal calibration sample[13].Our results also illustrate how an experiment is con-ducted on CAMEA-type instruments. The bandwidthof the magnetic excitations in MnF2 is roughly 6 meV.Since our analysers cover an energy range of 1.8 meV,four different incoming energies Ei were used. In addi-tion, the modular design of our analyser setup resultingin dark angles is to be considered. Thus, a full coverageof the scattering angles requires two different A4 settings,which are 4 degrees apart. For each Ei-A4 combinationan A3 sample rotation scan ranging from 0 to 150 degreesis performed in steps of 1 degree. The 8 different scanslead the dispersion shown in the left part of Fig. 16 forwhich a binning of n = 8 was used. In the left panel the3D data of the dispersions above different Bragg peaksare shown. The right panel reveals a cut through the 3Ddata along the principal along the principle axes (h,0,0),(-1,0,l) and (h,0,-1).0 Channel1 2 3 4 5 6 7 8E f [meV] Vanadium 3.177 3.370 3.551 3.757 4.006 4.292 4.612 4.963Full Simulation 3.199 3.374 3.570 3.789 4.037 4.316 4.634 4.999Back End Simulation 3.203 3.378 3.573 3.792 4.040 4.319 4.638 5.003FWHM Prismatic 5 [ µ eV] Vanadium 96 109 116 121 134 150 174 198Full Simulation 98 106 119 132 146 166 186 209Back End Simulation 57 64 70 78 87 96 111 124FWHM Prismatic 3 [ µ eV] Vanadium 109 123 127 131 145 165 188 212Full Simulation 105 114 126 139 156 176 197 222Back End Simulation 69 78 84 93 101 113 128 145FWHM Prismatic 1 [ µ eV] Vanadium 142 162 174 186 202 230 258 292Full Simulation 132 144 160 177 198 224 253 286Back End Simulation 104 114 126 141 159 181 202 235TABLE III. Comparison of final energy and resolution between Vanadium scan and McStas simulation. ( , , ) [ R L U ] −2.5−2.0−1.5−1.0−0.50.00.5 ( , , ) [ R L U ] −2.0 −1.5 −1.0 −0.5 0.0 0.5 E [ m e V ] -0.48-0.000.00 -0.74-0.000.00 -1.00-0.000.00 -1.00-0.000.33 -1.00-0.000.67 -1.00-0.001.00 -0.500.001.00 -0.000.001.00 HKL E [ m e V ] FIG. 16.
Top : 3D visualization of the measured dispersionrelations of MnF . Bottom : Cut along a line from (h,0,0),(-1,0,l) and (h,0,-1). Points in a distance of 0.05 1/A perpen-dicular to the cut were binned, while energies from 0.45 meVto 6.95 meV are binned into 66 equi-sized bins of 0.118 meV.
In just 28.5 hours the total data acquisition took place,measuring the spin wave dispersion in a large part of re-ciprocal space, covering all energies from 0.35 meV to6.95 meV. Despite the rather large sample, this full map-ping is not feasible with a standard TAS setup. Usingthe prismatic 8 pixels, a total of 9.5 mio individual datapoints are measured. If all of these points were to bemeasured on a standard TAS, only 0.01 s would be al-lowed per second, whereas on average each data point ismeasured for 73 seconds. A direct comparison is, how-ever, not completely possible as sensitivity of neutrondetection changes across CAMEA.The continuity and smoothness of the shown data con-firms that the CAMEA is working very well. It is onlypossible to have a consistent intensity level of the disper-sion and the background if all of the detector normal-ization steps work. Together with the regularity of thedispersion itself the use of geometry calculations to findA4 are deemed successful.We further note that the 2D cuts show the instrumentresolution of the inelastic signal, because it is knownthat the spin waves of MnF are known to be resolutionlimited[12]. In particular we point out the in and outof focus sides of the dispersion are clearly seen around(-1,0,0). (h,0,0) is out of focus, making it wide and lessintense than (-1,0,l).The high data quality allowed an extraction of the cou-pling constants. Since complete form of the magneticHamiltonian is already known[12], it was directly appliedto our data using the SpinW software package[14]. Theresults of these fits are plotted as red circles in Fig. 17.The obtained the exchange parameters are tabulated intable IV, where J and J describe the Heisenberg in-teractions between same site Mn and different site Mnrespectively. D d d denotes the anisotropy term. Our find-ings are close to the reported values of Okazaki et. al. [15].1 J [meV] J [meV] D d − d [meV]Current Findings 0.02681 0.15585 0.06710Tabulated Values[15] 0.028 0.152 0.091TABLE IV. Resulting coupling constants for the Hamiltoniandescribing MnF .FIG. 17. Fitted magnon positions (red circles) from Fig. 16overlayed with simulated intensities from SpinW. The fittedparameters are presented in table IV. We note that uncertain-ties are smaller than marker size. D. Currat-Axe spurion
A triple axis spectrometer exploits Bragg scatteringfrom the monochromator and analyzer crystals (typi-cally pyrolytic graphite for cold neutron instruments)to deteermine the inelastic response of a single crys-tal. Together with the sample three scattering pointsare present. Whenever two out of the three crystals(sample, monochromator and analyser) are in Bragg con-dition, a weak signal from the third may be detected.This constitutes 3 different cases; A Bragg reflection inthe monochromator and analyzer contributes to the elas-tic scattering. If the sample scattering angle acciden-tally corresponds to a Bragg reflection of the analyzeror the monochromator, incoherent scattering will resultin a spurious peak known as a Currat-Axe spurion[16].Since CAMEA-type instruments cover a large range ofscattering angles and energies transfers, Currat-Axe spu-rions are of widespread occurrence. Two spurions areproduced above each sample Bragg peak. An example ofthese is shown in Fig. 18. We note that while the spu-rion arising fom Bragg reflecion of the monochromatoris well visible, the one from the analyzer is suppressedsubstantially by the Be filter. −0.4−0.20.00.20.4 , , [ R L U ] −0.4−0.20.00.20.4 , , [ R L U ] −1.4 −1.2 −1.0 −0.8 −0.61, 0, 0 [RLU]−0.4−0.20.00.20.4 , , [ R L U ] -1.18-0.000.24 -1.09-0.000.12 -1.00-0.00-0.00 -0.91-0.00-0.12 -0.82-0.00-0.23 HKL E [ m e V ] -7.00E+00-6.80E+00-6.60E+00-6.40E+00-6.20E+00-6.00E+00-5.80E+00-5.60E+00 FIG. 18. Currat-Axe spurion in the MnF data set above (-1,0,0). Left : Constant energy planes at 0.6 meV, 1.0 meV,and 3.0 meV ± . meV, showing the presence of Currat-Axespurions below the dispersion cone. The one originating frommonochromator is shown in red and the one from the ana-lyzer in orange. Right : Q-E cut around (-1,0,0) along ˆ r = (-0.61,0,0.8), using a width of 0.075 1/Å. The chosen directioncoincides with the spurion. We note the multiple spurionsoriginating from the 4 different incoming energies of the dataset. Intensities are given on logarithmic scale. We note that a prediction of where Currat-Axe spu-rions appear in reciprocal space is straight forward andis implemented in the MJOLNIR[2] software. In Fig. 18we calculated them for the (H,0,L) plane of MnF . Thediscontinuities in the monochromator Currat-Axe arisefrom the 4 different incoming energies. V. CONCLUSION
The CAMEA secondary spectrometer has been in-stalled at the continuous spallation neutron source SINQof the Paul Scherrer Institut. The results obtained dur-ing the first commissioning phase substantiate our ex-pectations of the spectrometer concept in terms of us-ability, stability, and detection efficiency. The novelmulti-analyser-detector arrangement allows rapid map-ping of excitations without sacrificing the energy- andq-resolution of modern triple axis instruments. The in-strument will further benefit from an upgrade of the neu-tron guides and the replacement of the vertically focusingmonochromator by a doubly focusing one. Comparingthe instrument to the earlier RITA-2 instrument measur-ing 9 data points at a time, the CAMEA secondary spec-trometer increases this capacity by more than a factor700. We also report the energy resolution of all detectorpixels. The energy and intensity calibration have been in-corporated into the conversion algorithm of MJOLNIR[2]transforming raw data into reciprocal space.Raw data were generated at the Paul Scherrer Institut.Derived data supporting the findings of this study areavailable from the corresponding author upon reasonablerequest.2
ACKNOWLEDGEMENTS