The tracking detector of the FASER experiment

FASER is a new experiment designed to search for new light weakly-interacting long-lived particles (LLPs) and study high-energy neutrino interactions in the very forward region of the LHC collisions at CERN. The experimental apparatus is situated 480 m downstream of the ATLAS interaction-point aligned with the beam


Introduction
where the FASER detector is installed. This location enables FASER to search for new weakly-interacting long-lived particles (LLPs) with masses in the MeV to GeV range, produced from the copious inelastic pp scattering events at the LHC. The detector design has been validated against the benchmark physics process of a dark photon A decaying into a pair of oppositely charged particles, e.g, A → e + e − . Along with this process, FASER is also sensitive to several other new physics models presented in [2] which predict LLPs that travel to and decay in the FASER detector into pairs of Standard Model (SM) particles including photons. In addition to the LLP searches, the FASER location allows to study neutrino interactions of all flavors in an uncharted energy region [3] 1 . The experiment location greatly reduces background from the SM entering FASER from the LHC collisions, since such particles would have to traverse the LHC magnets, absorbers and 90 m of rock.
A sketch of the FASER detector with the coordinate system is shown in and is currently being commissioned towards data taking during LHC operation in 2022-24 (Run 3). As a decay volume for LLP searches, the detector has a 1.5 m-long and 0.55 T dipole magnet with a 10 cm radius aperture to 1 Notably, a 29-kg pilot emulsion detector was temporarily installed in this location during LHC running in 2018, which has been used to observe the first neutrino interaction candidates at a collider [4]. 2 Note, the figure shows some detector components that have not yet been installed. is not needed, the detector is designed to be able to measure charged particle momenta up to a few 100 GeV, with the performance determined by the relative alignment between the tracker stations. The tracker stations consist of silicon microstrip detectors with a pitch of 80 µm, which fully cover the aperture of the magnets; this paper will focus on details of these tracking detectors. The most downstream detector is a pre-shower scintillator station and an electromagnetic calorimeter with a depth of 2 and 25 radiation lengths (X 0 ), respectively, which will allow to discriminate electrons from muons and to measure the electromagnetic energy. A trigger signal for data taking is provided from all the scintillator stations and the calorimeter. Based on FLUKA [5,6] simulations and in situ measurements [7,8], a trigger rate of around 500 Hz is expected for a luminosity of 2 × 10 34 cm −2 s −1 . 3 This rate is dominated by high-energy muons penetrating through the 90 m of rock to get to FASER. Details of the trigger system are described in [9]. All components are integrated and have been commissioned 4 , as described in Section 5. To allow neutrino measurements at FASER, the detector will be augmented with an additional veto scintillator station, a 1.1-ton tungsten/emulsion de-tector that works as the target for the neutrino interactions, and an interface tracker. These elements will be installed in front of the FASER detector (described above) in late 2021. The interface tracker will enable tracks from a neutrino interaction in the emulsion detector to be matched to events in the tracker stations. This will allow the measurement of both the muon neutrino and muon anti-neutrino interaction cross sections by using the charge of the muon reconstructed in the FASER spectrometer. The interface tracker will also improve the background rejection and energy reconstruction for the neutrino analysis. The interface tracker has an identical design to the tracker stations used in the FASER spectrometer. FASER will therefore have four tracker stations in total, but this paper focuses on the three tracker stations used in the FASER spectrometer and already installed in TI12 in March 2021.
The layout of this paper is as follows. Section 2 presents the mechanical design of the tracker stations including details of the silicon strip detectors and on-detector electronics. The power supply, cooling, and data acquisition systems are outlined in Section 3. The interlock and detector control systems are summarized in Section 4. All components are integrated and have been commissioned 5 , as described in Section 5. Finally, conclusions and outlook are given in Section 6.

Tracker stations
The FASER tracker consists of 72 double-sided silicon microstrip modules arranged in three stations, each station being composed of three planes with eight modules per plane. Given the short timescale to build the entire tracker, spare barrel modules of the ATLAS Semiconductor Tracker (SCT) have been used.
The SCT is one of the three sub-systems composing the Inner Detector [10], the silicon tracker of the ATLAS experiment at the LHC [11]. 5 The results of the commissioning tests have been used to determine which modules were installed in which position in the detector to maximize the physics performance.

Silicon microstrip modules
The SCT barrel module [12]   precision coordinate (perpendicular to the strips) and ∼580 µm in the nonprecision coordinate (parallel to the strips) [12]. The flex hybrid, equipped with six ABCD3TA chips per side, is bridged over the sensors via a carboncarbon substrate. Two NTC thermistors (one per hybrid side) allow to monitor the module temperature. The baseboard, made of Thermal Pyrolytic Graphite (TPG) with an excellent in-plane thermal conductivity and low radiation length, provides the mechanical support to the sensors and allows to dissipate the heat generated by the FE electronics. The hybrid is attached to beryllia (BeO) facing plates located on the two ends of the TPG baseboard.
The modules used for the FASER tracker have been selected among the existing spares of the SCT barrel module production that was completed in 2004.
Since then, the modules were stored at CERN in individual sealed bags within a controlled environment. Some of these spare modules failed the initial quality assurance by the SCT collaboration, typically because there were more than 1% defective channels, high leakage current (above 4 µA) or early breakdown before 500 V (the maximum voltage expected after 10 years of operation in ATLAS). Given the much lower radiation levels expected in FASER compared to ATLAS 6 , the nominal operating voltage for the modules is set to 150 V for the entire detector lifetime. This bias voltage is large enough to ensure full depletion of the sensors, and it is not expected to be increased during the lifetime of the experiment due to the lack of cumulated radiation damage. Therefore, for FASER, out of modules operational up to 300 V, the ones with the lowest number of defects have been chosen. The IV curves have been measured at different stages of the tracker assembly (more details are given in Section 5). A tracker plane consists of an aluminum frame (AW-5083) holding eight SCT barrel modules as shown in Fig. 4. The modules are arranged by four on each side (front and back) of the frame to minimize non-sensitive material in the central region. To measure the momentum of the charged particles separated by the magnetic field, the modules are oriented with the strips perpendicular 6 A total ionizing dose less than 5 × 10 −3 Gy per year and a total fluence less than 5 × 10 7

Tracker plane
1-MeV-neutrons equivalent neq/cm 2 per year are estimated from simulations with FLUKA and confirmed with in-situ measurements in the TI12 tunnel [8]. The SCT barrel module is designed to maintain performance even after a total ionizing dose of 100 kGy and a total fluence of 2 × 10 14 1-MeV-neutrons equivalent neq/cm 2 [12].
to the y-axis (sensitive coordinate). The distance between modules (closest sensors along the out-of-plane direction) is 2.4 mm, and the active area overlap (in-plane, along the strip-length) is 2 mm. The overall active area in the tracker plane is 240 mm × 240 mm, which covers the 200 mm-diameter magnet aperture as illustrated in Fig. 5. The frame has a size of 320 mm × 320 mm × 31.5 mm. In order to minimize the material in front of the silicon detectors, the frame is cutout for most of the active area within the acceptance of the magnet aperture.
More details about the material distribution in the active area of a tracker station are given in Section 2.4. with respect to the plane reference system. All assembled planes were measured at the University of Geneva using a Mitutoyo CRYSTA-Apex S CNC coordinate measuring machine with automatic probe changer (Fig. 7). The precision of the metrology machine for in-plane and out-of-plane measurements is 5 µm and 10-15 µm, respectively. It is found that all frames are within the required tolerances (±20 µm) with respect to the CAD manufacturing drawings. A maximum deviation of 100 µm is found for the positioning of the SCT modules with respect to the CAD model (corresponding to a perfect alignment). This maximum deviation accounts for the combined effect of machining tolerances, SCT modules positioning errors and assembly precisions. The required precision of the alignment is O(50 µm). The initial alignment will be based on the metrology data, but can be improved and the stability monitored using track data during FASER operations.   In addition, LVDS receivers are placed to receive the operation signals for the ABCD3TA chips from the TRB. HV and Low-Voltage (LV) as well as their return lines are provided from a HV splitter board and a LV protection board to bias the sensors and power the ABCD3TA chips, respectively (see Section 3.1).

Pigtail and patch-panel
Every patch-panel is connected with the HV splitter board, biasing the sensors on the four modules with one HV channel. LV for the analog and digital circuits In addition, there are 5 V power lines on the patch-panel for the LVDS repeaters and receivers as well as a line to transfer an interlock signal generated by the TIM to the LV power supply.

Tracker stations
A tracker station is an assembly of three planes (including the patch-panels) as shown in Fig. 10. Each station is flushed with dry air to reduce the relative humidity ensuring safe operating conditions against dew point as discussed in Section 2.2. One inlet for the dry air is used for the whole station, while the other two are sealed by screws. All aluminium parts are post-treated with a trivalent chromium passivation 9 to prevent any corrosion by the corona effect that may occur after putting the frames in contact during the station assembly.
In addition, an O-ring sealing joint between frames provides a good tightness to keep the humidity inside the station as low as possible (typically about 1-2%).
The station is assembled from three planes only after the metrology and full commissioning of each individual plane is completed. The inter-plane alignment is done via two high-precision pins located in 5 mm-diameter H7 holes while the fixation between two adjacent planes is done via four M5 screws. The two end covers that close the station volume are made of 400 µm-thick carbon-fibre plates (standard T300 fibers). Without cables, the total weight of one station is about frame design. Over-heating of the SCT modules can cause problems in the mechanical integrity and alignment of the SCT modules due to the glass-transition of glues. The glue 10 used for the SCT module assembly should have the lowest glass-transition temperature around 35°C, which should be the absolute maximum temperature for the SCT modules. The safety scheme to ensure this is described in Section 4. Figure 11 shows the FEA of two planes in a station. The different simulation parameters have been set to match the testing conditions during the plane and station commissioning at CERN. In particular, the water temperature for cooling the tracker station has been fixed at 15°C, the water flow at 3 l · min −1 (for a heat transfer coefficient of water of 500 W · m −2 ), and the outside air convection at 23°C. The FEA gives a maximum temperature on the FE chips of ∼ 28°C, neglecting the temperature rise within the water channel due to the heat load (estimated to be +0.6°C for 3 l · min −1 ). The   Two different reference positions have been considered to estimate the material distribution along the z-axis: a central region, close to the geometrical center of the plane, and an edge region. Table 1 summarizes the material budget in each case. The central region corresponds to a particle traversing the least amount of material inside the station, i.e. six silicon sensors and two carbon-fibre covers that account for a total of 2.1% of a radiation length (X 0 ). The edge region is a worst-case position that corresponds to a particle traversing the six SCT modules (including sensors, TPG baseboard, flex hybrid with carbon-carbon bridge and readout ASICs), aluminum frames and station covers, accounting for a total of 21.5% X 0 . For a benchmark dark photon model (m A =100 MeV, = 10 −5 ) for dark photons that decay in the FASER magnet aperture, 70% will be in the low material central region of the tracker. Given the range of particle momentum expected in the experiment the contribution of multiple scattering from the traversed material is expected to be negligible. Total / station --2.1% 21.5% Table 1: Amount of material in X 0 in the active area of a tracker station for two regions: i) the central region with only the silicon sensor material and ii) the edge region. Details of the material in the SCT module are given in Table 8 of Ref. [12]. The numbers are calculated directly from the CAD description of the tracker station.

Tracker backbone
The  The backbone is the primary global mechanical structure of the tracker as it links together the three stations. It allows for an easy handling and transportation (Fig. 14a), and serves as a reference structure for the tracker alignment.
The backbone is then supported by the first and second short magnet cylinders via clamps (Figs. 14b and 14c). The clamp system is a small mechanism that is bolted onto the magnet cylinders and allows some angular movements of the backbone (and thus the tracker stations) with respect to the magnetic field.
This angular tuning is the only degree of freedom in view of the tracker station positioning during the survey by the CERN group. One clamp gives precise positioning along the z-axis by means of a locating pin, while the second one provides the alignment with respect to the x-axis with some dedicated slotted holes.

Powering, cooling and readout system
Since access to the FASER location is guaranteed only every two or three months during technical stops of the LHC, it is important that the system is  tor system is shown in Fig. 15, which also includes components for the FASER interlock and detector control system described in Section 4.

Power supply system
In order to provide HV and LV to the SCT modules described in Section 2.1, three 19-inch rack mountable crates called the MPOD LV/HV Power Supply System 11 are installed in TI12, which host 3 HV modules 12 and 18 LV modules 13 in total. For the patch-panels, the readout system (see Section 3.3) as well as the interlock and detector control system (see Section 4), one 19-inch rack mountable box (PSbox) is also installed, which holds fifteen 24V power supplies 14 .
Three types of printed circuit boards, the HV splitter board, the LV protection board and the 24V/6V board, were developed. These are mounted on the detector, directly above the tracker stations. The HV splitter board divides one HV channel into the four to supply the SCT modules connected to one patch-

Cooling system
The cooling system was designed and built by the CERN cooling and ventilation group (EN-CV). It consists of two air-cooled water chillers 16 : one circulating chilled water to the tracker stations and the second acting as a hot spare.
A manifold distributes the chilled water to each tracker station in parallel.
As shown in Fig. 16, the cooling system is mounted on a single frame together with all instrumentation. An additional water reservoir is also installed, which makes it possible to refill the water tank inside the chiller in an automatic manner. The cooling capability of each chiller is about 1.8 kW at a 15°C water outlet temperature with ∆T = 3°C between inlet and outlet temperature. Since one SCT module consumes 6 W, corresponding to an overall power consumption of 450 W for the full tracker, the cooling system is therefore sufficient to regulate the temperature of the tracker stations.
In case of a failure of the chiller in use for the tracker stations, the cooling circuit is re-routed by controlling valves to the spare chiller to take over the densation on the electronics due to the cooling. In case of a lack of dry air an alarm is triggered and the cooling and detector will be stopped by the hardware interlock system (see Section 4.3).

Readout system
One Tracker Readout Board (TRB) reads out eight SCT modules corresponding to one tracker plane. Therefore, a total of nine TRBs are used for the three tracker stations in the FASER spectrometer. The TRB consists of a GPIO board and an adapter card as shown in Fig. 17. Logic Board (TLB) [9]. The TLB, also a GPIO board, is the central trigger The software is written in C++ code and manages all operational procedures such as the calibration sequence as described Section 5.4 [9]. The firmware is

Interlock and Detector control system
The aim of the safety system of the FASER tracker is to protect the delicate silicon tracker modules from damage under all circumstances. The FASER tracker follows hereby the common approach of a multi-level protection system consisting of a high-level software-based detector monitoring called the Detector Control System (DCS), and a low-level hardware-based interlock system. The software system is capable of triggering automatic actions that can turn off individual detector components in a controlled way while the hardware interlock system turns off power supplies immediately and acts therefore as the last level of safety. Figure 19 gives an overview of the protection system of the FASER tracker.

Safety scheme
In the upper left corner, one of the two patch-panels of a tracker plane is sketched. Each patch-panel serves four silicon strip modules and each module is equipped with two temperature sensors (NTC-10k thermistor) that are used for temperature monitoring close to the readout part of the tracker module.
In addition, each patch-panel also connects to one temperature sensor (NTC-10k thermistor) that is thermally attached to the mechanical frame inside of the plane as the frame makes the thermal contact to the silicon strip modules In the following sections the specific hardware components of the tracker safety system are described in more detail. Figure 20 shows a picture of a TIM board as well as its corresponding block diagram. The core of the TIM is the AM335X micro-controller as well as three independent comparator based interlock circuits. The TIM can serve up to six tracker patch-panels via DB25 connectors each connector carrying the signal from one frame temperature sensor, 4 × 2 module temperature sensors, and one humidity sensor. There is one additional DB9 input that allows to connect up to 2 additional environmental temperature sensors. The latter is currently not used in the FASER trackers setup, but leaves flexibility for further monitoring in the future.

Tracker Interlock and Monitoring
All the mentioned sensors are routed to the microprocessor for digitization and monitoring. In addition, the frame temperature sensors of patch-panels that belong to the same tracker plane enter into one of the three comparator- Finally, the TIM features several interfaces for higher level communication: an Ethernet port for the communication with the DCS as well as several serial interfaces for debugging and optional connection to other devices.

MPOD Interlock Board
The

The DCS system
The DCS system is the central location where all monitoring data from the TIM boards as well as from the power supplies come together and is stored persistently into a database. Additionally, it provides high level controls of all the power supplies. The software is, therefore, capable of executing automatic actions in case the detector is leaving the operation parameter space indicated by one of the many available sensors. Due to the significantly larger number of sensor readings in the DCS system compared to the interlock system (see Section 4.2), a more sophisticated warning and protection scheme can be implemented. It should be noted however that this is not a safety system and does not replace the actual interlock system as it depends on a software process.
Details of the DCS system are described in [9]. Table 2 shows the thresholds values of temperature and humidity used for the operation of the tracker safety system. The thresholds for the automatic actions as well as the hardware interlock need to be chosen carefully such that the software-based automatic actions will step in before the actual hardware interlock in case of any abnormal temperature increase.

Limits for DCS automatic actions and hardware interlock
In order to verify these limits, a potentially destructive test was performed on a prototype tracker plane. All protection mechanisms were disabled, the cooling was stopped, and the plane remained fully powered. Figure 22 shows the thermal evolution measured by a representative frame (T frame ) and module temperature sensor DCS warning DCS automatic action hardware interlock was reached after about 9 minutes. The corresponding module temperature at that moment was measured to be T module = 32.0 • C and was therefore well below the required maximum module temperature of 35.0 • C. The threshold for the automatic DCS actions are adjusted in such a way that they are sufficiently far away from the normal operation point, but low enough in order still trigger before the hardware interlock under normal circumstances.

Tests during construction and commissioning of the FASER tracker
The FASER tracker elements were tested at each stage of the construction.
The electrical performance and behavior of the silicon sensors as a function of the applied bias voltage were investigated for the single SCT modules, individual tracker planes and full tracker stations on the surface before installation of the tracker stations into the FASER experimental site. In addition, metrology was performed for the layers and stations. After the installation, the performance was also tested using cosmic rays and random triggers deploying the central DAQ system of the FASER experiment. In this section, the test setup, procedure and results of the thermal and electrical tests are described.

Test setups during testing on surface
Three test setups were installed to perform the required measurements at the different stages, namely tests of a single module, plane and station.
The single module test was done with the readout system developed at Cambridge University [14]. It was used to evaluate the electrical performance of a single module. A chiller was used to cool down an environmental enclosed box to keep the module below 30 • C. The number of functional strips, the noise value and dependency on applied bias voltage (high-voltage behaviour) of the silicon sensors were verified and compared to the results in the module production of the ATLAS SCT detector [12].
The individual planes and stations were qualified on surface in test-stands which used equipment later used in the FASER experiment. Both the interlock and monitoring of the temperature, voltage and current were handled by the TIM unit (see Section 4.2). This allowed the DCS information to be monitored live and to be archived into a database for evaluating the detector performance.
The same powering and DAQ systems, cables and calibration software as used for the experiment were used during the testing on surface. This allows to compare the electrical performance at different stages of the assembly and installation as well as from different data taking periods in the tunnel. The surface commissioning with the components used in the real experiment made it possible to test their operation and investigate the long term behaviour early on.
The operation of the different tracker planes and stations was conducted with specific finite state machines which took the modified detector mapping on the surface compared to the one in the tunnel into account. Figure 23 shows a picture of the test setup with its main components for station commissioning. the test results were compared to the surface commissioning data. A general agreement was found as shown in more detail below.  As an example of the temperature stability of the tracker planes, Figure 25 shows the temperature measured with one of two NTC sensors on each module in plane 5 during a period of 24 hours covering one calibration sequence.

Results of thermal tests
The analogue and digital currents consumed by the ABCD3TA chips fluctuate during the calibration, causing a slight variation of the module temperature, however this never exceeds 29 • C. When the module is powered off, the temperature decreases to the coolant temperature of 15 • C. It is confirmed that the temperature was stable during nominal operation. In addition, the dew point inside the planes is kept around −40°C, which is well below the coolant temperature during the commissioning.

Calibration procedure
The readout of the SCT barrel modules is binary, meaning there is only a hit or no-hit information depending on whether the current pulse generated by a charged particle passing through the silicon sensor is above or below the threshold set. Since the analog information of the signal pulse is not recorded, a good calibration of the ABCD3TA chips on the SCT module is essential. The calibration of the chips mostly relies on so-called threshold scans (see Fig. 26), in which the discriminator threshold (differential voltage generated from an internal 8-bit DAC in the range from 0 to 640 mV) is varied in discrete steps and a set of well defined fixed-amplitude calibration charges is sent at each step.
The hit occupancy, defined as the fraction of injected signals above threshold, is computed for each readout channel at each threshold point. Since the signal amplitude is convoluted with Gaussian electronics noise, the hit occupancy does not follow an ideal step function along the threshold setting. The distribution behaves as the so-called s-curve. The threshold at which the occupancy is 50% is  The calibration procedure implemented for the FASER tracker largely follows that established by the ATLAS SCT collaboration [15], which has been extensively used during past years to perform the electrical characterization of the SCT modules. Typically, several tests / scans are run in sequence, and the chip parameters are updated along the sequence. A typical calibration sequence is listed below.
1. Mask-scan. This test aims at determining two sorts of main defects, i) dead / non-responsive channels, and ii) very noisy channels. This is achieved by setting a very low (high) threshold, then sending a number of trigger signals (without charge injection) to read the data stored in the digital pipelines, and finally checking how many channels are below (above) threshold. The strips with very low (high) occupancy at the low (high) threshold setting are identified as dead / non-responsive (noisy channels).
The identified defective channels are disabled by tagging them in a dedicated in-chip mask register. used. Using a linear fit to the vt 50 as a function of trim value, a channel is flagged as "trimmable" if its offset can be corrected with respect to a given threshold target. The trimming settings are those that correspond to the minimum TrimDac range and the minimum threshold target (within that range) for which a maximum number of channels are trimmable. This is done on a module-by-module basis to achieve a good threshold uniformity across all channels of a given module.

5.
ResponseCurve. With this test an accurate threshold-to-charge relation is obtained. Threshold scans are performed for ten input charges, from 0.5 fC to 8 fC. After the corresponding s-curve fits, for every channel the vt 50 as a function of the input charge is fit to a first-degree polynomial with an exponential correction term to account for small non-linearities at very low and high injected charges, τ = p 2 + p 0 /(1 + e −q/p1 ), where p 0 , p 1 and p 2 are the fit parameters, and τ and q are the threshold in mV and fC, respectively. The average parameters for each individual ASIC are obtained to determine the mV to fC conversion. 6. NoiseOccupancy. Although this test is formally not part of the chip calibration (as no calibration parameters are derived from it and it is not used to mask any additional strips), it is typically performed to assess the goodness of the above-mentioned procedure. The noise occupancy (NO) is defined as the probability for a strip to give rise to a hit only due to noise. This typically occurs when fluctuations at the discriminator input exceed its threshold. For the ATLAS SCT modules the noise occupancy per strip is specified to be less than 5 × 10 −4 at 1 fC threshold and the nominal operating temperature. The NO is determined by performing a threshold scan without any input charge. The number of triggers sent is increased progressively as the threshold is raised.

Results of electrical and calibration tests
The calibration scans described in Section 5.4 were repeated during individual plane commissioning, station commissioning, and commissioning in situ after installation in TI12 as described in Section 5.1. The aim of the calibration is to achieve uniform threshold distribution, high hit efficiency (> 99%) and low noise occupancy (< 5 × 10 −4 ) at the nominal operating threshold of 1 fC.
In this sub-section, electrical calibration results are shown and compared be-  In order to achieve a high hit efficiency, the number of masked strips -those identified as either dead or very noisy during the mask scan -in each module was carefully monitored during the tests for a single module and the plane and station commissioning. The modules with lower numbers of masked strips were selected to be mounted at one of the four inner module positions which are inside the central region of the magnet acceptance, as demonstrated in Figure 12. In addition, the station with the highest quality -that with lowest number of the masked strips in this region -was chosen to be placed at the front (upstream) of the spectrometer while the station with the lowest quality is located at the back (downstream). An electron and positron pair from the decay of a dark photon is most collimated at the upstream station, therefore, the station with the best performance was selected for that position. Table 3 shows the number of masked channels observed during the in situ commissioning after the installation. The four modules in the inner region have a maximum of 0.08% of masked channels.
Even in the outer region, this is less than ∼ 0.3%.

Station
Inner region Outer region  The gain measured in the three-point-gain measurements are shown in Fig. 28a.
The average over all the strips in all stations is 54 mV/fC, which is in good agreement with the ∼ 55 mV/fC expected from the module specification [12]. The dependence of the gain over a larger range of injected charges is tested in the response curve scan. Figure 28b shows the typical result of the response curve scan for a single strip.
After the trim scan, an even response between different channels in a module is obtained as shown in Fig. 29. The majority (99.9%) of the channels can be trimmed using the two lowest trim range settings, as expected for unirradiated modules, and an additional 0.05% of channels can be trimmed using one of the larger trim ranges. The remaining channels cannot be trimmed even at the largest trim range setting. It is important to keep the level of readout noise as low as possible to maintain low thresholds and accordingly realize high tracking efficiency. An estimate of the ENC at the discriminator input is obtained, using the three-point-gain measurement during the routine calibration scans described in Section 5.4. This provides a measurement of the output noise value with an injection charge of 2 fC. The distribution of ENC for all strips in the three stations are shown in Fig. 30a, whose mean value is 1454 ± 67 electrons. This is in good agreement with the ∼ 1500 electrons expected for unirradiated modules from the specifications [12].
The additional dedicated noise occupancy scans were also performed for a more direct measurement of the noise. The results for a single module are shown in Fig. 30b. Due to the large number of triggers required at high thresholds these take a considerably longer time than the three-point-gain measurement. The noise occupancy scans are, however, more sensitive to the tails of the noise distribution and to external sources such as a commonly increased noise on several where N and σ thr denote the normalization and ENC in fC, respectively. The chip-by-chip ENC evaluated from the noise occupancy scan and the three-point-gain measurement correlate well as shown in Fig. 30c, and are in good agreement with measurements of the ATLAS SCT modules [12]. Fig. 30d shows the noise occupancy at the nominal 1 fC thresholds using randomly triggered events during combined system runs in situ, which provides a further cross-check of the measured noise.
Over 99.7% of the strips in the tracker satisfy the performance criteria that the noise occupancy per channel is less than 5 × 10 −4 at the nominal 1 fC threshold.
The main reason for noisy strips is a non-linear behavior in the response curve.

Magnetic field test
One of the concerns for the installation of the FASER tracker in between the gaps of the permanent FASER magnets was the presence of the stray magnetic field during the lowering of the tracker planes. In order to exclude any negative impact on the performance or damage due to electromagnetic induction a test stand was set up that allowed to lower the FASER prototype tracker plane vertically in a controlled way at different distances to the aperture of one of the FASER magnets for testing purposes.
In a first step, the tracker plane was positioned 1 m away from the aperture (reference position) and a standard calibration and characterization sequence was run, i.e. the identification of noisy and non-responsive strips as well as the measurement of gain, threshold, electronics noise, and the sensor IV response.
At the reference position the magnetic stray field is negligible. In a second stage, the mechanical mounting structure was placed in a distance of 65 mm from the magnet aperture which corresponds roughly to the distance of the first tracker plane in the final FASER assembly. At this location, the magnetic field strength of the stray field amounts to up to 60 mT. The plane was lowered at a very low speed of about 0.03 m/s until it was finally centered in front of the magnet aperture. During the lowering process all cables (data, power, DCS) remained attached to the patch-panel of the plane, but were disconnected at the off-detector end. A second set of performance measurements was taken at this position. In a final step, a further reference measurement was taken at the reference position after the plane has been extracted in the reverse order with respect to the installation.
The number of noisy and non-responsive strips was found to be very consistent between the different measurement positions (maximally 2 strips of the difference in the prototype tracker plane). The same conclusion applies to the measured gain and noise, i.e., maximally ±0.3% and ±4% difference in the gain and noise of each module, respectively.
In conclusion, no sign for any damage or performance degradation could be found in this installation tests.

Cosmic ray test
In order to test the FASER tracker station as a full detector system including trigger and DAQ, a cosmic ray test stand was set up in the FASER surface laboratory. Figure 23 shows the setup in which the tracker station is placed horizontally in between two large trigger scintillators covering the full detector acceptance. Beyond the bare proof of the tracker functionality and its interplay with the larger FASER trigger/DAQ system, the collected cosmics data set is very valuable for the intra-station alignment in the offline reconstruction due to the absence of any stray magnetic field which will not be the case once integrated in between the FASER permanent magnets. The successful cosmic ray test concluded the tracker station commissioning on the surface.

Summary of commissioning
Three stations consisting of nine planes were successfully assembled and commissioned on the surface and subsequently in situ after installation. Overall a very good performance was found, which is comparable between surface plane commissioning and station commissioning and in the tunnel. The long term stability of full stations were additionally confirmed in long runs which lasted several weeks and was partially accompanied with data taking of cosmic ray events.

Conclusion and outlook
The FASER spectrometer was constructed with three tracker stations, comprising of nine planes of the silicon strip modules that were originally the spares for the ATLAS SCT barrel detector. All other parts of the tracker, including dedicated support frames, DCS and cooling system, readout electronics and services were newly developed for FASER. To ensure a precise alignment, within the mechanical tolerances, after the construction and assembly metrology of the tracker was performed for each plane and station.
The installed detector is fully operational and shows excellent performance, well within the specifications. The number of dead channels is less than ∼ 0.3% (including the modules in the outer region). The electrical performance of the detector was tested after calibration of the ABCD3TA readout chips at each stage of the construction. During the in situ commissioning of the final FASER setup, the average noise (ENC) was evaluated to be 1454 electrons, and over 99.7% of the the strips in the tracker satisfy the requirement of the noise hit occupancy < 5 × 10 −4 at 1 fC threshold. Long-term operation as part of the commissioning showed that all the SCT modules can be kept below the required maximum temperature of 35 • C with coolant temperature of 15 • C. In addition to standalone tests, the tracking detector was tested as part of the full FASER detector commissioning using the final TDAQ system, for example for combined cosmic ray data taking.
A new tracker station, identical to the three already installed in FASER, and called the "Interface tracker", will be installed between the, yet to be installed, emulsion detector and the FASER spectrometer. This Interface tracker will allow matching tracks between the emulsion detector and the FASER tracker, and enable to distinguish ν µ and ν µ interactions by measuring the charge of the produced muon in the FASER spectrometer. The Interface tracker construction and surface commissioning started in June in 2021, and the installation is planned for late-2021.
The FASER experiment will start physics data-taking in proton-proton collisions from the start of LHC Run 3 operations in 2022. The tracker will act as one of the crucial detectors to allow to search for new light long-lived neutral particles in FASER.

Acknowledgement
We thank the technical and administrative staff members at all FASER in-