Validation of the New Small Wheel Micromegas sectors for the phase 1 upgrade of the muon forward spectrometer of the ATLAS experiment at CERN

The ATLAS experiment is currently upgrading the first muon station in the high-rapidity region with the New Small Wheels (NSW), based on large-size resistive Micromegas (MM) technology and small-strip Thin Gap Chambers (sTGC). The NSW system is going to be installed in the ATLAS underground cavern during the LHC long shutdown 2 (2021) to enter in operation for Run3 (starting in the first months of 2022). 128 Micromegas quadruplets, each composed by four measurement layers two to three square meters in size, are needed to build the two New Small Wheels, covering a total active area of about 1280 m2. The construction of all MM modules, carried out in France, Germany, Italy, Russia and Greece, is completed. Their mechanical integration into sectors, the installation of on-detector services and electronics, for the first NSW is also completed, along with all validation and acceptance tests. The preparation of the second NSW is very well advanced. The advanced status of the project, in view of the imminent installation of the two NSW in ATLAS by the fall of 2021 will be reported. Micromegas detector construction and integration into sectors is presented, together with results obtained with cosmic rays data during the validation tests and the impressive steps of the wheel assembly completion.


ATLAS and the NSW
ATLAS [1] (figure 1 (left)) is one of the main experiments at the Large Hadron Collider (LHC) [2] and is undergoing the phase 1 upgrade in order to cope with the higher luminosity foreseen for the future LHC Runs. The old Small Wheels were the first muon spectrometer station in the forward region (End-Cap) with an angular coverage in pseudorapidity of 1.3 < |η| < 2.7 and were composed by Cathode Strip Chambers and Monitored Drift Tubes. An improvement in the performances was required in the NSW [3] both for the trigger and the track reconstruction in view of the increasing luminosity of LHC (5 − 7 × 10 34 cm −2 s −1 ) to maintain or even improve the performances of the detector.
Micromegas (MM) [4] and small Thin Gap Chambers (sTGCs) were chosen as fast detectors able to perform precision tracking (∼100 μm per plane [5] with an efficiency >90%) and to cope with the increasing background particle flux as the luminosity increases (up to 20 kHz/cm 2 ) while rejecting fake triggers.
The ATLAS MM detectors have a trapezoidal shape, to match with the wheel structure of the NSW as shown in figure 1 (center). Each of the 2 NSWs is composed of 16 sectors: 8 small (SM1 and SM2) and 8 large sectors (LM1 and LM2). The production is distributed between several industries and institutes: Italy (SM1) [6], Germany (SM2), France (LM1), Russia-Greece-CERN (LM2). Each sector of the NSW is a sandwich of 2 sTGC Wedges and a MM Double Wedge as shown in figure 1 (right). The NSW MM wedges are formed by 2 quadruplet detectors one after the other, type 1 detectors at lower radius and type 2 at larger radius as can be appreciated in figure 1 (right), with 8 printed circuit boards (PCBs) in total (5 from type 1 modules and 3 for type 2 modules). Any further distribution of this work must maintain attribution to the author(s) and the title of the work, journal citation and DOI.

Micromegas operating principles and NSW structure
The resistive Micromegas chambers are frontier Micro-Pattern Gas Detectors with a planar geometry as shown in figure 2 (left) operating in a gas mixture of Ar: CO 2 (93%:7%); studies ongoing on a new ternary gas mixture will be presented later. They have a 5 mm conversion (drift) gap, a floating mesh embedded in the drift panel structure and a 128 μm wide amplification gap between the read-out (RO) PCBs and the mesh (supported by insulating pillars of millimetric diameter). The PCBs are produced by industries, with 300 μm wide strips and a pitch of 425-450 μm. Resistive strips are superimposed to the copper signal strips to mitigate the intensity of discharges [7]. The detector, in this fashion, is really compact with a high electric field (∼45 kV cm −1 ) on a surface of O(m 2 ) in the amplification gap, high transparency and with fast ions evacuation (∼100 ns).
NSW MM were produced for the first time on large dimensions (O(m 2 )) and were constructed with some peculiarities with respect to the status of the art of the pre-esistent MM technology: • Screen printed resistive strips capacitively coupled to copper read-out strips, in order to cope with the high flux expected in view of the HL-LHC future Runs.
• Mesh at ground potential in order to allow for separation of the anode in separate HV sections, which was required by the fact that industries had limitations in dimensions of the PCBs.
• Mechanically 'floating' mesh, which is integrated in the drift panel structure and not embedded in the anodic structure (as it was for the bulk MM [4]). This is necessary for large area detectors and allows for chamber reopening in case of intervention.
Each MM chamber is a quadruplet formed by 5 stiff panels needed to form 4 gaps when coupled (figure 2 (right)): 2 RO panels, and 3 drift panels composed by cathode PCBs and the meshes. Two out of four layers have strips inclided by ±1.5°in order to reconstruct the 2nd coordinate (stereo layers), while the other two aim for precision coordinate reconstruction (eta layers). The mesh is grounded, the RO resistive strips are at ∼570 V and the  cathode at −300 V. Resistive strips are screen printed with equidistant interconnections to have uniform resistance across the pcb, with a design resistivity of ∼10 MΩ/cm. This peculiar structure of the ATLAS MM allows the detector to be re-opened for intervention since the mesh is not glued on top of the RO panel, and makes an easier construction procedure; each PCB is diveded into 2 HV sections, having in total 40 HV channels for the type 1 ATLAS MM chambers (5 PCBs per layer) and 24 HV channels for the type 2 ones (3 PCBs per layer).

MM construction: challenges and solutions
Mechanical requirements were so stringent and demanding: 36 μm of precision in η on positions of strips over meters was required on strip alignment on each layer and was reached by mechanical supports to the PCB during panel construction together with optical measurements (Rasnik technique [8]) of reference masks etched on the external side of PCBs. The technological transfer of RO PCBs production with extremely high quality (pillars shape, resistivity homogeneity, quality of the PCB edges) was one of the most delicate points since, during construction we faced a serious problem of HV stability, whose solution was affecting different aspects of the production.
The main issues and solutions were identified to be: • Residual ionic contamination of boards and panels from industrial processing and handling -addressed by improving the cleaning procedures; • Possible effects from mesh mechanical imperfections -addressed by implementing mesh polishing; • Clear correlation of currents with humidity -addressed by monitoring the internal humidity and increasing the gas flux; • Low and non-uniform resistance of the resistive layer (screen printing technique) and the strong dependence on the layout (design issue) -edge passivation [9].
As from the resistive circuit of the strips, the resistance is smaller at the edges. The mentioned issues were causing the fact that the resistance at the edges was actually lower than the sustainable limit, therefore passivation was producing an increasing of the miminum resistance at the edges. Having this in mind, and mainly with the passivation solution, we managed to address the HV issue. As shown in figure 3 there is a clear correlation between the minimum resistance and the sustained HV by each section as obtained from the tests on SM1 chambers before using the passivation technique; the passivation technique is also shown. In order to gain the most from all the benefits introduced, the number of HV channels has been increased by a factor 3 with respect to what originally programmed (roughly 1 HV channel per layer), in order to cope with weak sections and allowing us to run with only few% of sections not at nominal voltage.

MM double wedges integration and validation
Once at CERN, the MM detectors were mechanically assembled in Double Wedges (DWs), i.e. 2 detector Wedges (8 layers in total) and fully tested with cosmic rays data using a self-tracking method. The MM Double Wedge is the mechanical integration of two MM Wedges made of 4 layers each: the Wedge closest to the collision point is referred to as the IP side, the external one is the HO side. A schematic view of the 8 layers of the MM DW can be shown in figure 4. The reconstructed track is obtained from a linear interpolation of the clusters from 5 layers (different from the one under analysis), requiring at least 2 eta and 2 stereo layers. Two different planes of scintillator are used to trigger the events with a trigger rate of ∼120 Hz. The cluster on the analysed layer which is closer to the extrapolated position obtained from the reconstructed track on the layer itself, is chosen and used for the analysis if the residual (the difference between the cluster position and the extrapolated track position) is less than 5 mm. This value has been chosen far greater than the expected resolution of MM layers (∼100 μm) in order to take into account possible effects such as multiple scattering that could affect the resolution. The strips used in the analysis are those that belong to the selected cluster, with at most 1 hole allowed, charge greater than 0.4 fC, with the highest strip charge greater than 1.2 fC and the associated position is calculated using the centroid method (sum of the strip position weighted with the strip charges). The cluster is therefore formed by a number of adjacent strips with current above threshold, allowing maximum for 1 hole in between: the number of strips forming the cluster is referred to as the cluster width. Typically, at nominal voltage, the cluster is composed of ∼4 strips as shown in figure 6. Tracking hits on a single layer and efficiency results are shown in figure 5 together with a picture of the Cosmic Stand for the MM Double Wedges at CERN. The distribution of the hits on track (reconstructed) reflets the area covered by the scintillators, i.e. the fact that perpendicular tracks are more probable. For a single layer fully working (electronics and HV) we achieved efficiencies at the level of 98% and very good homogeneity. Studies of the MM performances can be done also as a function of the angle of the incident track (in order to reproduce the ATLAS conditions of the NSW) by selecting incident tracks with a defined range of angles (see section 5). In figure 6, some examples of performances are shown for some layers from different sectors of the Wheel A (which go from sector A01 to sector A16). Studies have been carried out performing a scan over the amplification voltages to clearly see the cluster width and cluster charge increase while reaching the efficiency plateaux as presented in figure 6. Results show also that, already at 560 V, the single-layer cluster efficiency is well above 90%. All the integrated Double Wedges have fulfilled the validation requirements of having a cluster efficiency greater than 99% with only 1 exception of a large Double Wedge whose chambers have been passivated and re-assembled at CERN to cope with the low resistivity of the resistive strips and are now fully working.

MM surface commissioning on the NSW
The Commissioning involves all the validation steps of the sectors installed on the wheel as shown in figure 7 and, at regime, was taking 2 full days after service installation. It involved the preparation of the MM DWs for the integration in a full NSW sector together with 2 sTGCs wedges, then all the preparatory steps were performed to have the sector ready for the installation. Checks on the integrity of fibers, T-sensors, cables and gas pipes were performed together with checks on the grounding. Each sector ready for the installation was then installed on the wheel and services were connected as first step. After that, the validation was proceeding through all the checks involving the Low Voltage connections and polarity for every cable going from the detector to the power supply. Then the cooling system was tested together with the gas tightness of the MM detectors before starting with the final steps: HV validation and tests involving the electronics, with specific data taking to test the DAQ and trigger system functionalities and to ensure the electronics itself in view of the phase 1 and phase 2 upgrade needs.
During the wheel A surface commissioning, a huge problem has been faced regarding the level of noise on the electronics. We therefore studied the efficiency of the MM detectors as a function of the threshold to be set on the electronics in order to overcome the noise level. The efficiency studies shown in figure 8 have been performed on a MMFE8 basis (MM electronics components 1 ) for different angles of the reconstructed incident tracks as already discussed in section 4. As can be seen, the same behaviour is shown independently of the angle: the efficiency drops if the threshold set on the electronics to overcome the noise is too high. Figure 9 shows that  the noise found at the beginning was reaching peaks of 6000 Equivalent Noise Charge (ENC), corresponding to a ∼50 mV threshold to be set on our electronic channels. This level of noise was not affordable at all and was leading the cluster reconstruction efficiency down to ∼85%-90% depending on the angle range of the reconstructed track as reported in figure 8. After critical months of investigation of the noise sources and studies, the main issues have been identified in the grounding quality and in the power distribution, therefore grounding has been reinforced with additional braids on the detector bases and RO pcbs, the power distribution has been refurbished adding an output common mode filter and a capacitive filter to cut the common mode noise (2-10 MHz) and the T-sensors power modules grounding has been improved. Similar issue have been observed by sTGC, with similar solutions.
At the end of the NSW commissioning all the sectors have been tested and successfully validated, leading to the installation of both the NSWs in the ATLAS cavern for the LHC Run 3.   Isobutane (Iso, iC 4 H 10 ) is a quencher, therefore it helps in stabilizing the current and dumping the spikes, allowing also to run at significantly lower amplification voltages with equal gain and efficiency (typically −60 V on the HV) with respect to the binary gas mixture as shown in figure 10 [11]. Spikes are defined as instantaneous currents above a certain threshold, in this case 100 nA (this threshold is set empirically). It has been observed that bad HV-sections behave better while adding Isobutane, improving the sparking picture and allowing for better performances. Currents have been measured with a Keithley 2470 sourcemeter and results are shown in figure 11 [9] for the same good HV section at nominal voltage in Ar: CO 2 (93: 7) and in Ar: CO 2 : iC 4 H 10 (93: 5: 2); the nominal voltage is 570 V for the former and 520 V for the latter. It can be seen that the frequency of sparks in a 0.2 nA arbitrary binning for  the current is greater in the case of the binary gas mixture. The improvement in terms of spark frequency, even in the case of a good behaving section is clear and quantified also by the number of sparks above a 5 nA threshold, which, in the case of the ternary gas mixture, are one order of magnitude less. Studies are ongoing at GIF++ (Gamma Irradiation Facility using a 137 Cs source) at CERN, in parallel with the NSW installation in ATLAS, in order to investigate ageing effects. The first results show that all the problems observed in few HV sections becoming resistive (chambers have been retested with cosmic rays data, re-opened, repaired and re-assembled) were due to weak points indepentent of the gas mixture. The few points emerged have been anyway fixed by reopening of the chambers, confirming it as a great opportunity.

Conclusions
The NSW, consisting of new MM and sTGC detectors will be working as part of the ATLAS experiment as starting from the Run 3 of LHC. MM will be used for the precision tracking in the NSWs.
With all the knowledge acquired in the past years we managed to address the main issues affecting the MM detector construction and integration (mostly HV issues and noise issues) in time for the installation.
Studies with Isobutane enriched gas mixture show promising results in terms of improving the performances of the chambers and we aim to use it from start.
It has been an incredible work done by the collaboration during such hard times and an impressive achievement that has been possible thanks to the commitment and dedicated effort of hundreds of people.

Data availability statement
The data cannot be made publicly available upon publication because they are not available in a format that is sufficiently accessible or reusable by other researchers. The data that support the findings of this study are available upon reasonable request from the authors and belong to CERN on behalf of the ATLAS Collaboration (Creative Commons Attribution 4.0 License https://creativecommons.org/licenses/by/4.0).