GEM Tracking Detectors for COMPASS


COMPASS is a two-stage magnetic spectrometer, built for the investigation of the gluon and quark structure of nucleons and the spectroscopy of hadrons using high-intensity muon and hadron beams from CERN's SPS. After the muon beam program will be completed, experiments with a hadron beam of 2⋅107 particles/s are foreseen from 2008 onward to perform spectroscopy of mesons and baryons in the light quark sector. For these experiments the tracking of charged particles in the beam region requires fast detectors providing good resolution in space and time in order to disentangle pile-up and multi-track events. Further demands of the high hadron flux density to the detectors are radiation hardness and minimal mass in order to avoid secondary interactions. Based on the experience with large-area triple-GEM tracking detectors in COMPASS, a set of triple-GEM beam trackers with combined pixel readout in the central region and 2-D strip readout in the periphery has been proposed (“PixelGEM”). After a successful prototype test in high-intensity muon and hadron beams in 2006, a total of five detectors is being built and tested for 2008.

The GEM Principle

In 1968 the invention of the Multi Wire Proportional Chamber (MWPC) revolutionised particle detectors. Invented and developed by Georges Charpak, Nobel Prize laureate in 1992, its basic concept is simple: Thin anode wires are stretched between two cathode foils at high voltage, each wire acts as a proportional counter and amplifies the ionisation tracks, released by charged particles crossing the gas in the chamber.

But in 1997 Fabio Sauli, inspired by the work of Anton Oed on Micro-Strip Gas-Chambers (MSCG), introduced the Gas Electron Multiplier (GEM), by turning around the concept of MWPCs. Instead of anode wires, holes, etched in metal coated polymer-foils, are used for amplification in GEM detectors.

Still MWPCs and Drift Chambers (DC) are covering large areas and wide angels in spectrometers, but MicroMegas and GEMs, both Micro Pattern Gas Detectors (MPGD) seems to be a good choice for Small Area Tracking, as shown at COMPASS.

The GEM foil, short GEM, is a thin insulating polymer foil, coated on both sides with a very thin metal layer. The whole plane is perforated with a large number of circular holes. A standard-design GEM consists of a 50 µm thick Kapton foil covered on both sides with a 5 µm copper layer. The hole diameter in the copper (D) is 75 µm, the centre-to-centre distance (P) between the holes is 140 µm . The GEMs are produced in the CERN workshop with the help of photolithographic methods. The pattern of the holes is first engraved in the metal surface afterwards etched with a Kapton-specific solvent from both sides. This leads to the double conical shape of the holes, an important factor for the operation of GEMs. The Kapton tips prevent shortcuts between the two electrodes by operating voltages (∼400V). The holes have an inner diameter (d), from tip to tip, of 65 µm, slightly smaller than D.

Upon the application of a potential difference between the two metal electrodes of the GEM foil, a high energetic field is generated inside the holes. For a voltage difference of ΔUGEM ≈ 400 V, fields of 50 kV/cm are reached. If the GEM is placed in a parallel drift field, one gets a single stage gas amplifier. The electron is guided by the field lines into a hole of the foil. Due to the high electric field an avalanche multiplication occurs. The produced charge cloud follows the drift field lines as well and can be collected or further amplified. With a single foil gains above 103 can be reached.

The figure below shows the electric field lines in the GEM hole region calculated with MAXWELL. It is obvious that most of the field lines from the upper side enter the hole and exit on the lower side. But some field lines also end on the Kapton surface leading to a deposit of electrons around the Kapton tips. This process stops when the Kapton is saturated. This negative charge at the tips increases the electric field in the hole and therefore also the gain of the detector up to 30%. This phenomenon is called charging up. Due to the fact that the charging up is very fast (a few minutes), it is no problem for measurements, although the gain varies at the beginning.

Discharges, Ion Feedback and Transfer Coefficient/Electron Transperancy

Discharge probability, the ion-feedback and the electron transparency are very important features for the operation and the design of micro-pattern gaseous detectors. The discharge probability has to kept as small as possible, as the occurring high currents might damage the foils or the readout electronics. The production of positive ions in the amplification process is very serious. Since they are collected very slowly by the electrodes and generate a build-up of positive charge which modifies the electric fields. Therefore the influence of ions has to be limited. But at the same time the electron transparency has to be guaranteed. For COMPASS GEM detectors the main emphasis was placed on the reduction of discharges.

Extensive studies where performed to understand the discharge behaviour of GEM based detectors. While the discharges can be occur as a breakdown in the gas volume, also sparks in the holes of a GEM foil might happen. The probability of sparks depends on the potential (ΔUGEM) between the two electrodes of the foil. Tests with an α-source show that the probability is rather zero below 500 V and increases exponentially by higher voltages.

The possibility to cascade several GEM foils is one of the big advantages of the GEM technology. GEM stacks with up to four foils have been built and investigated. One result was the fact, that the discharge probability can be reduced to a minimum if several GEM foils (amplification stages) are combined. The reason for this is that they can operate at lower voltages. The figure below shows a drawing of a single and double GEM and introduces some general definitions.

In multiple GEM detectors discharges occur in the last foil, where the avalanche and the corresponding charge are biggest. Studies with different stack geometries (single, double and triple GEM) point out the dependency between detector gain and discharge probability. The results are presented in the second figure below. Based on these studies a triple stack geometry was chosen for COMPASS GEM detectors.

The purpose of GEM detector is the collection of an amplified electron cloud on the readout. Therefore one has to get rid of the slowly moving positive ions. In the drift volume the ions, created by primary ionisation, are released by the drift foil. But ions, created in the holes by avalanche processes, can disturb the electric field in the holes. In the best case they are guided to the upper side of the foils and do not enter the drift volume. If a big number of ions reach the readout, they induce a noise signal. The ion feedback can be defined as the ratio of the current induced at the drift foil by the ions moving in the drift field and the current induced by a electron charge cloud at the readout.

The fraction of negative charge which is guided from the drift volume into the GEM holes is called electron transparency. Just like the ion feedback it depends strongly on the field strength of the surrounded gas volume and on the GEM voltages.

Realization of a Triple-GEM Detector for COMPASS

COMPASS GEM detectors are triple GEM detectors, comprising three foils as a multiplication stack. The choice, using three foils for one detector, minimises the risk of discharges, necessary in the regions where the GEM detectors are placed, and ensures a high gain for an efficient detection of MIPs.


For COMPASS GEM detectors amplification foils with the standard geometry were chosen: on a 50 µm thick Kapton carrier plane a 5 µm copper layer is coated on both sides. The holes are of a double conical shape, the hole diameter in the metal is 75 µm and 65 µm at the tip. The dimension of the foil is 310 mm x 310 mm, the maximum possible size due to the production limitations. To avoid damages on the foils, caused by shortcuts in the holes, the available energy is reduced by segmenting the foil in twelve parallel sectors on the top-side. In order not to disturb the efficiency of the detector, the sector boundaries, areas without holes and no metal surface, has to be as small as possible, for COMPASS foils a width of 200 µm is usual, although some detectors have sector boundaries of 500 µm, because of manufacturing mistakes.

Additionally to the twelve sectors there is a central circular region, called beam killer or dead zone, separately powered by the 600 µm wide central boundary, independent from the other sectors. While nominal beam conditions, the centre is always off, otherwise the detector gets “blind”, due to the high occupancies, caused by the unscattered beam. For alignment runs with low intensity beams, the centre can be switched on.

The figure below shows a typical COMPASS GEM foil, the twelve sectors and the beam killer are visible, as well as the 14 conductor strips, powering the different sectors, the centre region and the unsegmented back side of the foil.

The 2D Readout Plane

The GEM detectors are able to read out both coordinates at the same time. This is realised by a Printed Circuit Board with two layers of 768 perpendicular coppers strips at a pitch of 400 µm, separated by 50 µm thick Kapton. To provide an equal sharing of the collected charge, the width of the strips is different. While the upper strips have a width of 80 µm, the width of the bottom strips is 340 µm (the readout of the first detectors have a width of 350 µm, but it turned out, that these strip width leads to short cuts between single strips). To the side is a schematic drawing o f the readout.

The strips lead out of the detector gas volume and are connected to the readout electronic via wire bonding. Due to diffusion the charge cloud collected on the readout board is bigger than the strip width (≈ 3.5 x pitch) and a weighting method is used for calculate the exact track position in two dimensions. The group of strips, being hit of the same charge cloud, are called cluster. A detailed discussion about the characteristics of these clusters, can be found in, as well as a solution for track reconstruction and analysis in the COMPASS GEM detectors.

An important point, having an effect on the noise of the front-end readout electronics, is the capacity of the strips. Every single strip versus the other readout coordinate acts as a plate capacitor. With the permittivity ε=3.9 of Kapton and an area of 2.27⋅10-1cm2, this capacitance is 15.7 pF.

The Complete Detector

The whole detector is glued between two honeycomb support plates of 3mm thickness. Both plates have a round hole in the centre to reduce material in the beam region. The hole in the top honeycomb is smaller (35 mm diameter versus 50 mm in the bottom honeycomb) to stabilise the drift foil, which is glued on it. The space of 3mm between the cathode and the first GEM foil is defined by a 3mm thick fibreglas frame, including the gas inlet. The distances between the different GEM foils and between the last foil and the readout plane is given by 2mm thick spacer grids, glued between the foils as a support for the GEMs to withstand the elctromagnetic force between the electrodes without bending. The ground honeycomb plate on which the readout plane is glued, has a width of 500 x 500 mm2, also including the gas outlet. The figure gives an insight of the structure of a COMPASS triple GEM detector, all lengths are in millimetres.

The Detector Gas

Depending on the amplification voltage, it is in principle possible to achieve avalanche multiplication in any gas or gas mixture. For the gas choice several criteria has to be taken into account, for the COMPASS GEMs as well as for other gas detectors.

Low working voltages are an important point by the layout of detectors. Due to the shell configuration of noble gas atoms, avalanche multiplication happens at much lower electrical fields, compared with complex molecules, making noble gases the main component of most gas detectors. For the detection of MIPs a noticeable primary ionisation and charge in the drift field is necessary. The specific ionisation increases with the atomic number and, as Krypton and Xenon are very expensive, therefore Argon is the filling gas used for COMPASS GEMs.

As a consequence of the primary ionisation and the avalanche amplification a lot of ions and excited atoms remain in the detector. The ion are neutralised at a cathode (at GEMs, the drift foil or the upper side of GEM foils) whereby a photon can be emitted. Exiting atoms also return to ground state by emitting an Auger electron or also a photon. The generated photons in the detector can hit the metal surface and free an electron via photoelectric effect. This electron can start a new amplification process in the detector and tamper the signal. Polyatomic molecules, called quencher, absorb the photons and release the gained energy by rotations or vibrations. In COMPASS GEMs CO2 is used as quencher gas, since it is not flammable and not polymerising.

Extensive ages studies were performed with the gas filling and the detectors,but still no ageing was observed.

Design of the PixelGEM Detector

The electron signal emerging from the triple GEM amplification stack of the PixelGEM detector is read out by a 100 µm thin, Kapton-based flexible printed circuit foil with three conductive layers, carrying both the pixel and the strip structure, and the signal lines from the pixel region. Pixels of 1 x 1mm2 size have been chosen since they are expected to yield cluster sizes larger than one, thus allowing to improve the spatial resolution by applying clustering algorithms. With a total of 32 x 32 pixels the beam region is fully covered while at the same time it is still technologically feasible to route the signals from the pixels on a single layer to the front-end electronics mounted 15 cm away from the active area. Surrounding the central square of pixels, a strip readout with 400 μm pitch has been realised on the same thin film printed circuit so that the complete active area of the detector amounts to 10 x 10 cm2. A total of 2048 channels are read out through 16 APV25 preamplifier/shaper ASICs with an analog pipeline with a capacity of 192 25 ns-spaced samples. Upon trigger, three samples are forwarded to digitisation and further processing, so that pulse shape analysis allows to determine the time of a hit relative to the trigger with a precision of <10 ns. For r<1.5 cm, i.e. in the centre of the detector, the amount of detector material was reduced to a total thickness of 0.4% radiation length (X0) and 0.09% interaction length (λI) respectively. Utilisation of GEM foils with Cu layer thickness of 1 µm instead of 5 µm is being investigated and may result in an even lower thickness of 0.2% X0 and 0.08% λI respectively.

Prototype Test

The picture on the left shows a top view of the full-size detector prototype: The detector is mounted on a light honeycomb sandwich structure serving as base panel and carrying the high voltage distribution, the front-end electronics, the readout foil, and the GEM stack. It was characterised in the laboratory using various ionisation sources, and then successfully tested in the 2006 COMPASS high-intensity muon beam at a flux density of about 5⋅104 µ/mm2/s. A different test making use of a 190 GeV/c pion beam focused at the position of the PixelGEM achieved a local intensity of the same order of magnitude. For the 2008 hadron beam a similar flux density is expected, although a greater area will be illuminated. First analysis shows a spatial resolution of ∼100 µm and an average cluster size of ∼3 in the pixel region. No electrical instabilities or discharges were observed during these tests, making this type of detector a promising candidate for a low-mass, radiation-hard beam tracker in high-rate hadron beams.