Recent Performance Studies of the GEM-based TPC Readout (DESY Module)
Ties Behnke, Ralf Diener, Ulrich Einhaus, Uwe Krämer, Paul Malek, Oliver Schäfer, Mengqing Wu
TTalk presented at the International Workshop on Future Linear Colliders(LCWS2019), Sendai, Japan, October 30th 2019.
Recent Performance Studies of theGEM-based TPC Readout(DESY Module)
Ties Behnke, Ralf Diener, Ulrich Einhaus, Uwe Krämer,Paul Malek, Oliver Schäfer, Mengqing Wu
Deutsches Elektronen-Synchrotron, DESY
For the International Large Detector (ILD) at the planned InternationalLinear Collider (ILC) a Time Projection Chamber (TPC) is foreseen as themain tracking detector. To achieve the required point resolution, Micro-Pattern Gaseous Detectors (MPGD) will be used in the amplification stage.A readout module using a stack of three Gas Electron Multipliers (GEM) forgas amplification was developed at DESY and tested at the DESY II TestBeam Facility.After introducing the readout module and the infrastructure at the testbeam facility, the performance related to single point and double-hit resol-ution of three of these modules is presented. This is followed by results onthe particle identification capabilities of the system, using the specific energyloss d E/ d x , and simulation studies, aimed to investigate and quantify theimpact of high granularity on d E/ d x resolution.In addition, a new and improved TPC field cage and the LYCORIS Large-Area Silicon-Strip Telescope for the test beam are described. The LYCORISbeam telescope is foreseen to provide a precise reference of the particle tra-jectory to validate the momentum resolution measured with a large TPCprototype. For this purpose, it is being installed and tested at the test beamfacility within the so-called PCMAG (Persistent Current Magnet).1 a r X i v : . [ phy s i c s . i n s - d e t ] J un ontents d E/ d x Resolution . . . . . . . . . . . . . . . . . . . . . 6 d E/ d x resolution 8 Introduction
This project is part of the design effort for the International Large Detector (ILD) atthe future International Linear Collider experiment (ILC) [1]. The FLC-TPC groupat DESY is part of the LCTPC collaboration [2], which is driving the efforts of de-veloping a TPC for the ILD. Different readout and amplification technologies basedon Micro-Pattern Gaseous Detectors (MPGD) are currently studied within the collab-oration. FLC-TPC is committed to designing a self-supporting readout composed ofGEMs [3] mounted on thin ceramic grids [4].
In order to maximize the sensitive area while minimizing the gaps and the materialbudget, the DESY GridGEM module has been developed. The mechanical base of themodule is an aluminium back frame that is used to mount the module in the TPCprototype end plate. Onto this a printed circuit board (PCB) is glued, which containsthe readout pad plane. This structure also provides the gas tightness of the module.The pad plane contains 4832 readout pads with a pitch of 1.26 mm arranged in 28concentric rows with a pitch of 5.85 mm in radial direction. On top of the pad boardthe amplification structure is built. It consists of three GEMs supported by 1 mm highalumina-ceramic frames. Additional frames are used as spacers between the GEM foilsto define transfer and induction gaps, as shown in Figure 1a. A fully assembled moduleis shown in Figure 1b. Since the frame bars are only ∼ (a) Explosion view of the GEM module. (b)
Picture of an assembled GEM module.
Figure 1:
The DESY GridGEM module. .2 The LCTPC Setup at the DESY II Test Beam Facility The LCTPC readout prototypes are tested in an electron beam at the DESY II TestBeam Facility [7]. The beam line delivers electrons in the momentum range from 1 GeV/cto 6 GeV/c at rates from O (kHz) to ∼
100 Hz, respectively.The setup consists of a large TPC prototype (LPTPC) and a superconducting solen-oid magnet, the so-called Persistent Current Magnet (PCMAG), shown in Figure 2a.A coincidence signal from four scintillator planes placed in the beam path in front ofthe PCMAG is used as the trigger signal for the main setup. Slow control is imple-mented using the Distributed Object-Oriented Control System (DOOCS). The solenoidproduces a magnetic field of up to 1 T. It is mounted on a movable stage installed in theexperimental area T24/1 of the test beam. The stage can be translated vertically andhorizontally perpendicular to the beam axis and rotated around the vertical axis. Addi-tionally, the TPC can be rotated around its axis inside the magnet. This allows to scanthe beam over the full volume of the TPC and to change both the polar and azimuthalangle of the beam relative to the TPC. The usable inner diameter of the PCMAG is85 cm. Due to a lightweight construction without a field return yoke, the magnet wallcontributes only about 20 % of a radiation length X [8].The LPTPC has a maximum drift length of about 570 mm and an inner diameter of75 cm. The anode end plate contains slots for 7 modules, arranged in three staggeredrows. Since all module slots are of identical shape, the rows are not concentric. Unusedslots can be filled with dummy modules for a uniform termination of the drift field. Thisis shown in Figure 2b, where the active modules can be distinguished by their morereflective surface. (a) The PCMAG. (b)
The LPTPC anode end-plate [9].
Figure 2:
The PCMAG solenoid and the LPTPC setup at the DESY II Test Beam Facility. .3 Established System Performance In two beam test periods in 2013 and 2016, respectively, the point resolution of thesystem has been determined. Generally good agreement was found between the data ofboth tests. A small discrepancy was found in the rϕ -resolution at small drift distances,as shown in Figure 3a, visible as a larger intrinsic resolution σ rϕ, of the system and asmaller transverse diffusion constant D t in the newer data. Various systematic effectshave been studied, including gas pressure and temperature, gas composition and con-tamination as well as settings of high voltages and electric and magnetic fields. Withinreasonable fluctuations, none of these effects can explain the observed discrepancy andfurther investigations are required.Using the measurement data, the point resolution can be extrapolated to the con-ditions of the ILD TPC. Since the transverse diffusion is reduced due to the highermagnetic field of 3.5 T, the expected diffusion constant is simulated using Magboltz [11].Figure 3b shows the resulting rϕ -resolution for the 2013 data set. The orange linerepresents the direct extrapolation of the measurement, including the measured electronattachment rate. Since the attachment rate is expected to be negligible in the ILD TPC,due to much lower gas contamination, for the green line the attachment rate was set tozero. In this case both data sets result in a resolution around σ rϕ = 100 µ m at the fulldrift length of the ILD TPC, fulfilling the requirement set in [1].A related quantity is the double-hit resolution, which is important in the case of close-by or partially overlapping tracks. Here, an algorithm was developed that first identifiespotential double-hit candidates based on preliminary track fits. It then performs fits tothe charge distribution of both a single-hit and a double-hit hypothesis. If the double-hit hypothesis performs significantly better than the single-hit hypothesis, the chargedistribution is split into two hits based on the fit. The blue dots in Figure 4 show the (a) Comparison of measurements. (b)
Extrapolation to the ILD.
Figure 3:
The transverse resolution σ rϕ (a) as measured in the two beam test campaigns.The markers represent the data and the lines show fits of the expected behaviouras described in [10]. (b) An extrapolation to the size and magnetic field of the ILDTPC. The colored bands show the σ confidence intervals. igure 4: The two-hit separation efficiency versus the distance between the respective tracks.The three different marker color represent different steps of the algorithm. Red: noattempted hit separation; green: intermediate step; blue: full algorithm. two-hit separation efficiency of this algorithm versus the distance between the two recon-structed tracks at the location of the hits. The other colors represent intermediate stepsof the process. The solid lines represent fits of an error function to the respective data.The efficiency reaches 50 % at distances around and below 2 mm, slightly depending onthe drift distance due to diffusion. This fulfills the requirement specified in [1]. d E/ d x Resolution
The average energy loss on a track is calculated from the charge distribution of all hitsthat were assigned to this track by a track finding algorithm [12, 13]. On average,each track in the large TPC prototype contains about 55 hits that are considered validfor the d E/ d x calculation. The mean d E/ d x per track in the LPTPC is shown inFigure 5a. The relative resolution is extracted from a Gaussian fit as σ divided bythe mean. To achieve a stable estimate of the average energy loss from the Landau-like primary ionization distribution the mean of a truncated distribution is calculated,where only a certain fraction of the lowest charge hits is taken into account. As shown inFigure 5b, a shallow optimum of the resolution is found around a truncation fraction of75 %. For tracks extrapolated to the full ILD TPC as described below the same optimumis found. Using the optimal truncation fraction the resolution for the measured tracksis σ d E/ d x = (8 . ± .
14) % .The final goal of this analysis is to estimate the d E/ d x resolution of the full size ILDTPC. Therefore, the results need to be extrapolated to the greater track length, i.e. thegreater number of possible hits. Naively, one could expect the resolution to scale withthe number of hits N as / √ N but experience of previous experiments shows, that itrather follows a power-law σ d E/ d x ∝ N − k with an exponent k in the range of 0.43 to 0.48, depending on experiment and d E/ d x calculation methods [14]. This is caused by the Laundau-like shape of the probability6 a) Track d E/ d x distribution. (b) Optimization of the truncated mean.
Figure 5:
The determination of the d E/ d x resolution. (a) The distribution of the truncatedmean of d E/ d x per track. The red line represents the Gaussian fit used to extractthe resolution. (b) The d E/ d x resolution as a function of the truncation fraction.The blue marks show the resolution for the measured LPTPC tracks and the greenmarks an extrapolation to the full ILD TPC. Figure 6: d E/ d x resolution obtained using pseudo-tracks of various length. The red linerepresents the power-law fit described in the text. d E/ d x value for each pseudo-track is then calculated as above. Samples of pseudo-tracks with various numbers ofhits are created and the resolution calculated for each. Figure 6 shows the resultingresolution versus the chosen track length. The power-law fit returns an exponent of k = 0 . ± . . The d E/ d x resolution at 220 hits as calculated from the fit is equal to σ d E/ d x = (4 . ± .
1) % , fulfilling the goal given in [1]. d E/ d x resolution Simulation studies have been performed in order to investigate and quantify the impactof high granularity on the d E/ d x resolution. Traditionally, the energy loss is computed by summing all the electrons generated fromionization by the incident particle. The number of electrons produced in the ionizationprocess is distributed according to a Landau distribution, which has a long tail towardslarge values. The relatively large width of the distribution is reflected in a degradationof the correlation between the measured energy and the momentum of the particle. Analternative method consists in counting the number of ionizing interactions of the in-cident particles. This follows a Poisson distribution with a significantly smaller width,compared to the Landau, therefore providing a better correlation and particle identific-ation power [17]. A comparison between the two methods has been performed studyingthe pion-kaon separation power as a function of the pad size, as shown in Figure 7, wherethe stars and the circles represent the simulation performed using the charge summationmethod and the cluster counting method, respectively. They are compared to resultsfrom test beam measurements represented by the blue squares. Below a pad size ofabout 500 µm the cluster counting method performs better than the charge summationmethod.
In order to exploit the cluster counting capability, a high granularity readout system isneeded. The ROPPERI (Readout Of a Pad Plane with ElectRonics designed for pIxels)system represents a possibility to combine high granularity, flexibility and large anodecoverage of the pad-plane readout [18]. 8 igure 7:
Pion-kaon separation power as a function of the pad size. Blue squares representresults from test beam measurements; stars and circles represent the simulationsusing the charge summation and cluster counting methods, respectively.
The idea of the ROPPERI board implementation is sketched in Figure 8. GEMs areused for amplification, small pads on a PCB form the anode and they are read out bya Timepix ASIC, which has a matrix of ×
256 = 65536 pixels with a 55 µm pitch.The connections from the pads are routed through the PCB to the ASIC which is bumpbonded to the PCB surface. The Timepix power and communication pads are on thesame side of the ASIC as the pixels, which makes it necessary to also connect them viabump bonds back into the PCB, where they are routed to a macroscopic connector plug(VHDCI).A prototype board had been designed to study different pad sizes and line lengths,and the influence of the capacitance on the signal to noise ratio. Several boards havebeen produced out of which seven have been bump bonded in different tries. Out ofthese seven boards, only three worked completely. During the data taking, differentthresholds have been used in runs with 100 to 200 frames. The bump bonding testsshowed issues connected to temperature during and after the bump bonding. Due totemperature difference and differing expansion coefficients of the materials of the chip
Figure 8:
The ROPPERI board. · to 2 · signal electronsarriving on a pad after the GEM amplification, a signal to noise ratio higher than 10can be estimated. DESY is working on the construction of a new large prototype TPC field cage. A highprecision TPC field cage for the ILD TPC has to guarantee a low material budget, ahigh voltage stability and high mechanical precision. Since the current large prototypefield cage, shown in Figure 9a, does not fully meet the requirements on the mechanicalprecision [20, 21], it was decided to build a new one. This field cage is build in the DESYworkshop, so it also represents an opportunity to gain in-house experience in the designand construction of such a composite structure. The prototype is supposed to fit intothe PCMAG, therefore its dimensions are restricted by the magnet geometry, as was thecase for the previous field cage. The outer diameter is foreseen to be 77 cm leaving agap of 4 cm to the inner wall of the magnet, which has a diameter of 85 cm.As stated above, the 61 cm long field cage should be lightweight and at the sametime mechanically very stable. Therefore, its structure is made of composite materials,already used in the construction of previous prototype field cages. Figure 9b shows asketch of the wall structure, which is estimated to contribute 1.31 % of a radiation length X . An optimized design of field strips and mirror strips, implemented on a double-sidedcopper-coated polyimide foil, is used to generate a homogeneous field inside the cage. A (a) The current prototype field cage. (b)
Sketch of the structure of the current field cagewall [19].
Figure 9:
The existing large prototype TPC field cage. a) (b)Figure 10: Schematic view of cuts through the field strip foils perpendicular to the field strips.(a) shows the current and (b) the new design. The field strips are depicted in black,the polyimide foil in red, the white boxes represent the resistors. Figures from [21]. resistor chain linearly degrades the voltage along the length of the field cage, creating aconstant field. At the time of the construction of the current field cage, no producer wasfound to manufacture the 61 cm ×
226 cm large, two-sided field strip foil in one piece.Therefore, it was produced from two about 30 cm wide foils which were connected duringthe construction. Now, the production of such a large foil in one piece is possible and aset of foils is being produced at the CERN workshop.After the construction of the current field cage, the design studies of the field striplayout have been carried on and a simpler design delivering a comparable field qualitywas found. In the new design, each mirror strip spans over two gaps so only half of theresistors and vias are needed compared to the previous design. Details can be foundin [21]. Figure 10 shows sketches of the cross sections of the old and new designs of thefield strips.
Figure 11:
Sketch of the functionality of the LYCORIS telescope (here denoted by “strip tele-scope”) installed within the PCMAG solenoid together with a large TPC prototypeunder test. igure 12: The LYCORIS telescope mounted in the PCMAG solenoid together with a Mimosatelescope.
The LYCORIS Large Area Silicon Strip Telescope [22, 23] is foreseen to be installedwithin the PCMAG at the DESY II Test Beam Facility. The system requirementsdescribed here are for the use case of a TPC within the PCMAG. The telescope isdesigned to provide a precise reference measurement of the particle trajectory, as shownin the sketch in Figure 11. These reference measurements can be used to study andcorrect for potential inhomogeneities of the electric field within the TPC volume, whichotherwise limit the achievable momentum resolution.To meet these criteria, the telescope needs to provide a spatial single point resolutionof better than σ bend = 10 µ m along the bending direction within the PCMAG and aresolution better than σ drift = 1 mm along the drift axis of the TPC.Hybrid-Less silicon strip sensors, designed by SLAC, are used for the LYCORIS tele-scope. They are characterized by an active area of 10 cm ×
10 cm with a strip pitchof 25 µm and contribute 0.3 % of a radiation length. The sensors are read out by anintegrated digital readout chip called KPiX, also designed at SLAC, which is directlybump bonded onto the sensors and routed to the strips within the silicon sensor.The telescope has been fully assembled and successfully tested in several test beamcampaigns, also in combination with EUDET-type Mimosa pixel telescopes [24]. Apicture of the LYCORIS telescope, used in combination with a Mimosa telescope insidethe PCMAG solenoid is shown in Figure 12. The system is undergoing final adjustmentsto be integrated into the infrastructure of the DESY II Test Beam Facility.12
Summary
The DESY GridGEM modules have been shown to be able to fulfill the requirementsset in [1] for several important parameters: point resolution, double-hit resolution and d E/ d x resolution. Simulation studies have shown that the d E/ d x resolution may beimproved with an optimized pad size of O (500 µ m) . The ROPPERI system indicates apossible way to achieve such readouts with very high granularity.A new, improved field cage for the large TPC prototype is in preparation. The Lycorisbeam telescope has been fully assembled and was successfully tested. Some more work isneeded to enable integration into the infrastructure of the DESY II Test Beam Facility. Figure 2b is used with permission from the LCTPC Collaboration as authors of thearticle “A time projection chamber with GEM-based readout”, Nuclear Instruments andMethods in Physics Research A 856 (2017) 109–118, Elsevier, c (cid:13)
References [1] T. Behnke, J. E. Brau et al., eds.
The International Linear Collider . TechnicalDesign Report. Vol. 4:
Detectors . Linear Collider Collaboration, 2013. arXiv: . url : .[2] Homepage of the LCTPC collaboration . url : .[3] F. Sauli. ‘GEM: A new concept for electron amplification in gas detectors’. In: Nuclear Instruments and Methods in Physics Research. Section A: Accelerators,Spectrometers, Detectors and Associated Equipment doi : .134] T. Behnke, R. Diener et al. ‘A Novel Self-supporting GEM-based AmplificationStructure for a Time Projection Chamber at the ILC’. In: Journal of Instru-mentation doi : . arXiv: .[5] R. Esteve Bosch, A. Jimenez de Parga et al. ‘The ALTRO chip: A 16-channelA/D converter and digital processor for gas detectors’. In: IEEE Transactions onNuclear Science doi : .[6] L. Jönsson and U. Mjörnmark. Front-end electronics and data acquisition for theLCTPC . Eudet-Memo 2007-53. EUDET, 2007. url : .[7] R. Diener, J. Dreyling-Eschweiler et al. ‘The DESY II test beam facility’. In: Nuclear Instruments and Methods in Physics Research. Section A: Accelerators,Spectrometers, Detectors and Associated Equipment
922 (2019), pp. 265–286. doi : . arXiv: .[8] A. Yamamoto, K. Anraku et al. ‘Balloon-borne experiment with a superconductingsolenoidal magnet spectrometer’. In: Advances in Space Research
14 (1994), p. 2. doi : .[9] D. Attié, T. Behnke et al. ‘A Time projection chamber with GEM-Based readout’.In: Nuclear Instruments and Methods in Physics Research. Section A: Accelerators,Spectrometers, Detectors and Associated Equipment
856 (2017), pp. 109–118. doi : . arXiv: .[10] F. J. Müller. ‘Development of a Triple GEM Readout Module for a Time ProjectionChamber & Measurement Accuracies of Hadronic Higgs Branching Fractions in νν H at a 350 GeV ILC’. Dissertation, Universität Hamburg, 2016. DoctoralDissertation. Hamburg: Universität Hamburg, 2016, p. 245. doi : .[11] S. Biagi. ‘Monte Carlo simulation of electron drift and diffusion in counting gasesunder the influence of electric and magnetic fields’. In: Nuclear Instruments andMethods in Physics Research. Section A: Accelerators, Spectrometers, Detectorsand Associated Equipment doi :
10 . 1016 / S0168 -9002(98)01233-9 .[12] J. Abernathy, K. Dehmelt et al.
MarlinTPC: A Marlin based common TPC soft-ware framework for the LC-TPC Collaboration.
LC Note LC-TOOL-2007-001.2007. url : http://bib-pubdb1.desy.de/record/83142 .[13] C. Kleinwort. A Track Finding Method for a TPC Based on fast Hough Transform-ation . LC Note LC-TOOL-2014-006. Deutsches Elektronen-Synchrotron, DESY,2014. url : http://bib-pubdb1.desy.de/record/193046 .[14] W. Blum, W. Riegler and L. Rolandi. Particle Detection with Drift Chambers .2nd ed. Particle Acceleration and Detection. Springer, 2008. doi : . 1415] T. Behnke and F. Gaede. ‘Software for the international linear collider: Simulationand reconstruction frameworks’. In: Pramana doi : .[16] F. Gaede. ‘Marlin and LCCD: Software tools for the ILC’. In: Nuclear Instru-ments and Methods in Physics Research. Section A: Accelerators, Spectrometers,Detectors and Associated Equipment
559 (2006), pp. 177–180. doi : .[17] M. Hauschild. ‘dE/dx and Particle ID Performance with Cluster Counting’. In: In-ternational Linear Collider Workshop. ILC-ECFA and GDE Joint Meeting (Valen-cia, Spain, 6th–13th Nov. 2006). 2006. url : https://agenda.linearcollider.org/event/1049/contributions/506/ .[18] U.Einhaus, J.Kaminski and M.Caselle. ‘ROPPERI - A TPC readout with GEMs,pads and Timepix’. In: International Workshop on Future Linear Colliders. (Stras-bourg, France, 23rd–27th Oct. 2017). 2018. arXiv: .[19] P. Schade and J. Kaminski. ‘A large TPC prototype for a linear collider detector’.In: Nuclear Instruments and Methods in Physics Research. Section A: Accelerators,Spectrometers, Detectors and Associated Equipment doi : .[20] T. Behnke, K. Dehmelt et al. ‘A lightweight field cage for a large TPC prototypefor the ILC’. In: Journal of Instrumentation doi : .[21] P. Schade. ‘Development and Construction of a Large TPC Prototype for theILC and Study of τ Polarisation in ˜ τ Decays with the ILD Detector’. DoctoralDissertation. Universität Hamburg, 2009. doi : .[22] M. Wu, M. Breidenbach et al. ‘Lycoris: A large area beam telescope based onhybrid-less strip silicon sensors’. In: Nuclear Science Symposium and MedicalImaging Conference. (Sydney, Australia, 10th–17th Nov. 2018). IEEE, 2018. doi : .[23] M. Wu, M. Breidenbach et al. ‘Development of a large active area beam tele-scope based on the SiD microstrip sensor’. In: Nuclear Instruments and Methodsin Physics Research. Section A: Accelerators, Spectrometers, Detectors and Asso-ciated Equipment (2019), p. 162864. doi : .[24] H. Jansen, S. Spannagel et al. ‘Performance of the EUDET-type beam telescopes’.In: EPJ Techniques and Instrumentation doi :10.1140/epjti/s40485-016-0033-2