The Pixel Detector of the ATLAS experiment for the Run2 at the Large Hadron Collider

The instantaneous luminosity of LHC will be upgraded to about $2\times 10^{34}$ cm^-2s^-1 after LS2 which is planned in 2018. The large amount of the hit data from the Pixel Detector will become a serious issue for the front-end electronics in the operation after LS2 since it was designed to work at the instantaneous luminosity up to $1\times 10^{34}$ cm^-2s^-1. Especially, the data transmission speed and buffer size of the MCC are not enough to cope with the data obtained at the instantaneous luminosity of $2\times 10^{34}$ cm^-2s^-1. If, however, a new pixel layer is installed inside the inner most layer of the Pixel detector, B-Layer, it can recover the efficiency drop of the Pixel Detector and improve the performance. For that reason, a new pixel layer (IBL) [4] was installed inside B-Layer during LS1. In the following, when comparing the technologies of the IBL with the pixel detector used for Run-1, the latter will be referred as the Pixel Detector.

The modules of the IBL adopt two different sensor technologies: n⁺-in-n planar sensors and 3D sensors. The main characteristics of these sensors are summarized in Table 1. The pixel size of the sensors is $50\times 250$ $\mu$m² which is 60% of the pixel size used for the Pixel Detector. The IBL has about 12 million pixels in total. The thickness of the planar and 3D sensors is 200 $\mu$m and 230 $\mu$m, respectively. The sensors were designed to work at a radiation dose up to $5\times 10^{15}$ n_eq which is the expected dose by the end of LHC Phase-1 operation around 2022.

n⁺-in-n planar sensor is a well developed technology and it is the one used for the Pixel Detector. n⁺-in-n planar sensor uses n-type bulk, and n⁺ implants and p⁺ implants are put as the electrodes at the sensor surface and the side of the bump-bonding with a front-end chip to apply the high voltage (HV) and ground, respectively.

The edge around the guard-ring is inactive in planar sensor. The sensors used for the Pixel Detector have an inactive region of 1100 $\mu$m from the edge, therefore, Pixel modules are placed with overlaps along the beam axis to compensate this inactive region. Since the space of the radial axis ($r$) is limited in the case of the IBL, it is impossible to overlap the modules. For that reason, planar sensors were developed with a smaller inactive region of 200 $\mu$m, extending the pixels at the sensor edge under the guard-ring [5].

In 3D sensors, n⁺ and p⁺ implants are embedded vertically into the p-type bulk as the electrodes to apply the HV and ground, respectively. Since the electrodes are put closely with a distance of 50 $\mu$m, the 3D sensor can be fully depleted with much lower HV, compared to a planar sensor. For 3D sensors used for the IBL, 6 V is enough for the full depletion before the irradiation. It was designed to have 200 $\mu$m inactive region at the sensor edge. The IBL is the first large scale application of this new technology.

A new front-end chip (FE-I4B chip) was developed for the IBL. The FE-I4B chip was produced with IBM 130 nm technology, and $80\times 336$ pixels are embedded per chip. It is designed to have a radiation hardness up to 250 Mrad. In FE-I4B chip, hit data in each pixel are stored locally, and the pixel matrix is organized in four pixel digital readout units which provide local hit processing and distribute a buffer to store hit data until receiving a trigger. The signals from the sensor are amplified and digitized in the analog and digital circuits respectively, and finally the charge information is output with 4-bit value of TOT (Time-Over-Threshold).

The cost of the bump-bonding is not proportional to the chip size but to the number of chips. For that reason, the front-end chip with large size reduces the cost of the bump-bonding. The FE-I4B chip was, therefore, designed to have a large area of $20.2\times 18.8$ mm². The size is 4.5 times larger than that of the FE-I3 chip which is used for the Pixel Detector.

The sensors and FE-I4B chips are connected with bump-bonding by using AgSn, and then the flexible PCB is glued on the sensor side to make a module. Two FE-I4B chips are attached on a planar sensor and one chip is used for a 3D sensor, therefore, the size of a planar module is two times larger than that of a 3D module.

Figure 3 shows the number of the produced modules and the fraction of the bad modules [3]. At the beginning of the production (L1 in Fig. 3), open and merged bumps were found in a high rate of the modules. One of the reasons of this issue was to use too much flux during the flip-chip process. The problem was solved after adopting flux-free flip-chip process. The total production yield was 75% for the planar module and 62% for the 3D module after solving the bump issues.

Figure 4 shows a schematic view of the IBL stave. A bare IBL stave is made of carbon foam with the size of $74.8\times 1.88$ cm² and its cross-section has a triangular shape. Inside a IBL bare stave, a cooling pipe with 1.5 mm diameter is inserted, which is made of titanium. The planarity of the $r$-direction is required to be within 350 $\mu$m to put the staves in the limited space between the beam-pipe and the B-Layer of the Pixel Detector. The material budget of the stave is only 0.6% ($X/X_{0}$) including the cooling pipe.

Two flexible PCBs (stave flex) are glued on the two sides of $z$-direction of the bare stave to provide the electrical lines. In the stave flex, the low voltage (LV) lines for the FE-I4B chips are made of aluminum to reduce the material budget.

On each stave, 12 planar modules are mounted in the middle region in $z$-direction covering 75% area and 8 3D modules in the edge region covering 25% area. As described in Sec. 3.2, there is not enough space to put the modules with overlap in $z$-direction for the IBL, so that the inactive region of $z$-direction must be reduced as much as possible. For that reason, the modules were mounted with a gap less than 350 $\mu$m along $z$-direction. The stave flex and modules were glued together, and then all the electrical lines were connected with wire bonding.

During the production of the staves, Al(OH)₃ was found on the bonding pad on 13 staves, which comes from corrosion of bonding wires made of aluminum. Al(OH)₃ can be produced with DI water, and besides, the corrosion process is accelerated with halogen elements like chlorine and fluorine. The water came from condensation during thermal cycling where temperature was changed from $-40$ to $40$ degrees to investigate thermal tolerance of the staves. Chlorine and fluorine were also found on the bonding pad, and they seem to accelerate the corrosion process. The source of chlorine and fluorine, however, was not identified eventually. To avoid the corrosion, the humidity control was improved in the staves testing.

Once the staves were produced, electrical functionality tests were performed to investigate their performance in detail. Any possible damage of the wire bonding and module surface was checked by optical inspection with a microscope. The quality of the sensors was confirmed with Current-Voltage (I-V) measurement. A breakdown voltage above 80 V and 20 V was required for the planar sensor and the 3D sensor, respectively. The functionality of the FE-I4B chips was checked by configuring the chips and reading data from them, including the tuning of the chip parameters. Finally, source scans with ²⁴¹Am and ⁹⁰Sr were performed to test the overall functionality of modules. In these tests, it was confirmed that IBL can be operated with a threshold value of 1500 electrons with noise values below 200 electrons. Figure 5 shows the number of the bad pixels in the 14 staves used for the IBL [6]. The fraction of bad pixels is required to be below 1% in this test, and all the staves well satisfied this requirement.

The integration of IBL started on February, 2014. The beam pipe used in Run-1 was replaced by a new one with an inner radius reduced from 29 mm to 23.5 mm. The IBL was integrated around the new beam pipe and inserted into the ATLAS detector together with it.

Before the integration, the cooling pipes in the staves were extended to 7 m in length by brazing. The cooling is one of the critical parts in the IBL: the sensors must be cooled down to $-40$ degrees due to the high radiation damage (The radiation dose on the sensors is expected to be $5\times 10^{15}$ n_eq at the maximum). Although C₃F₈ has been used for the Pixel Detector whose cooling capability is up to $-15$ degrees, its cooling power is not enough for the IBL. For that reason, CO₂ evaporative cooling is adapted for the IBL, since it has higher cooling efficiency which allows to reduce the diameter of the cooling pipe and then the material.

After brazing the cooling pipes, the 14 staves were put on the support structure on the IBL Support Tube (IST) which is located around the beam pipe to integrate the staves. Then, the service cables, about 5 m long, were connected at each edge of the staves to provide the data lines, HV and LV. Two or three staves were integrated on the IST every week, and the integration finished in two months in March 2014.

After the integration, the test system was connected with the service cables, and an electrical functionality test was performed through the service cables to find any damage of the detector during the integration. In this test, one shorted line and disconnected line were found in the service cables, and they were fixed by replacing the cables. Except for that, no damage and degradation of the performance were found.

The IBL was transported to the ATLAS experimental hall at 90 m underground on May 5, 2014 and installed into the ATLAS detector on May 7. After connecting the cooling pipes and service cables for the powering and environmental monitors, the performance of the IBL was tested again with the test system. In this test, the final DCS system was used for powering and reading the environmental monitors. The test system was connected with the service cables of the data lines to read-out the modules on the staves. Figure 6 shows the difference of the noise values between this test and the stave quality assurance test performed before integration of the staves [7]. In this test, it was confirmed that 100% of the sensors and FE-I4B chips is operational. Besides, there was no significant change of the performance from the stave quality assurance test.