First Results on 3D Pixel Sensors Interconnected to the RD53A Readout Chip after Irradiation to $1\times$$10^{16}$neq cm$^{-2}$

TL;DR

This paper reports on the performance of 3D pixel sensors interconnected with the RD53A readout chip, tested before and after high-fluence irradiation, demonstrating promising hit detection efficiency for HL-LHC upgrades.

Contribution

First experimental results on 3D pixel sensors with RD53A chip after extreme irradiation, advancing detector technology for HL-LHC upgrades.

Findings

Hit detection efficiency around 97% after irradiation

Successful integration of 3D sensors with RD53A chip

Validation of sensor performance post-irradiation

Abstract

Results obtained with 3D columnar pixel sensors bump-bonded to the RD53A prototype readout chip are reported. The interconnected modules have been tested in a hadron beam before and after irradiation to a fluence of about $1\times$$10^{16}$neq cm$^{-2}$ (1MeV equivalent neutrons). All presented results are part of the CMS R&D activities in view of the pixel detector upgrade for the High Luminosity phase of the LHC at CERN (HL-LHC). A preliminary analysis of the collected data shows hit detection efficiencies around 97% measured after proton irradiation.

Tables

Table 1. Table 1 : Hit detection efficiency summary table.

3D Pixel-RD53A Linear FE	$25\times 100\mu m^2$	$50\times 50\mu m^2$
Before irradiation	97.3%	98.6%
After irradiation	96.6%	97.5%

Full text

1 HL-LHC Requirements and 3D Pixels

Pixel detectors in the innermost layers of the HL-LHC experiments will have to survive up to a fluence which can exceed 2\times$$10^{16} neq cm*-2*, while preserving high tracking efficiency [1]. The total active sensor thickness should be about 100 to 150 $\mu$m in order to keep both the bias voltage and the power dissipation after irradiation to a manageable level, while at the same time allowing for a reasonable amount of collected charge to reach full hit detection efficiency. Moreover, the radiation damage reduces the effective drift distance of charge carriers because of charge trapping, so it is not useful, in the case of planar sensors, to increase the thickness beyond the above limits. The 3D pixels, where charge carriers have to travel distances much lower than the sensor thickness (only 35$\mu$m for a $50\times 50\,\mu{\rm m}^{2}$ pixel pitch independently of the sensor thickness which is the driving parameter for planar pixels), are hence very good candidates to satisfy all of the above requirements.

2 The 3D Columnar Pixel Sensors

The 3D sensors [2] were fabricated at the FBK foundry in Trento; they were developed in a collaboration program with INFN (Istituto Nazionale di Fisica Nucleare, Italy). The substrates selected are p-type Si-Si Direct Wafer Bond (DWB) or SOI (Silicon On Insulator). The handle wafer is 500 $\mu$m thick low resistivity Czochralski (CZ). FBK active devices are implanted on a Float Zone (FZ), high resistivity (>3000 Ohm cm), 130 $\mu$m thick wafer. Columnar electrodes of both $p^{+}$ and $n^{+}$ type are etched by Deep Reactive Ion Etching (DRIE) in the wafer using a top-side only process. There are two different pixel cell pitches: $50\times 50\,\mu{\rm m}^{2}$ and $25\times 100\,\mu{\rm m}^{2}$, the latter having one or two collecting electrodes per cell (2E). Two examples of pixel cells are shown in figure 1.

Both cell sizes are presently under evaluation in CMS for the inner layers of the upgrade pixel detectors for HL-LHC. After fabrication pixel sensor wafers were processed for UBM (Under Bump Metalization), thinned down to $200\,\mu{\rm m}$ total thickness, diced and bump-bonded to RD53A prototype chips [3] at IZM (Berlin, Germany). The RD53A chip has 76800 readout channels (400 rows and 192 columns with a bump pad picth of $50\times 50\,\mu{\rm m}^{2}$) and measures $20.0\times 11.8\,mm^{2}$. The pixel sensor bonded to the readout chip needs eventually to be glued and wirebonded onto an adapter card in order to be tested; these units will be referred to as modules in the following. All results presented here were obtained with the Linear Front-End in the central zone of RD53A (136 columns wide, from 128 to 263).

3 Irradiations

Irradiations were performed at the CERN IRRAD facility in 2018 in a high intensity 24 GeV/c proton beam, which has a FWHM of 12 mm in x and y directions. The target fluence was 1\times$$10^{16} neq cm*-2*. Modules were tilted on the IRRAD beam at an angle of 55∘ in order to homogeneously irradiate the $20\times 12$ mm2 sensor and readout chip area; the corresponding total ionizing dose was 6 MGy for 1.65\times$$10^{16} protons cm*-2*. Cross checks were performed in order to establish the effective fluence integrated on the modules; visual inspections after irradiation and data analysis are showing that the modules were displaced with respect to irradiation beam axis by a few millimeters and that the irradiation is less uniform than expected. We estimated that the nominal fluence was reached in about half of the linear FE. All results shown here are based on the nominal requested equivalent fluence. In figure 2, the modules mounted on the adapter cards are shown. The two CMS modules were labeled w3x3y2 and w91x1y3 to identify the production wafer and the sensor position on the wafer itself.

4 Data Taking and Results

Hybridized modules were tested in two test beam experiments in CERN’s North Area H6B before and after irradiation in July and October 2018. Irradiated modules were kept cold at temperatures between $-20{\rm{}^{\circ}C}$ and $-30{\rm{}^{\circ}C}$ using dry ice bricks. The temperature was monitored via PT1000 sensors located close to the backside of the module and via NTC resistors soldered on the adapter card; these sensors gave consistent measurements. The tuning of the readout chip parameters was done on the beam-line for all the tested modules, targeting low thresholds and noise, having at most 1.5% masked pixels because of noisy channels. For the irradiated modules the average signal threshold was set to about 1400 electrons, with a noise value of 105 electrons for non-masked pixels, as shown in figure 3 for a $25\times 100\,\mu{\rm m}^{2}$ pixel size 3D module.

The color scale for the threshold distribution represents the 4-bit DAC value used for the trimming of each individual pixel response. The 3D modules before irradiation reached hit detection efficiencies above 98.5%, for perpendicular incident tracks, already at moderate HV bias. In our analysis dead pixels are excluded from the hit efficiency calculation. After irradiation, a bias voltage of at least 120V is needed to reach high hit efficiency. A comparison of hit efficiency before and after irradiation is shown in figure 4 for $25\times 100\,\mu{\rm m}^{2}$ and figure 5 for $50\times 50\,\mu{\rm m}^{2}$ pixel size modules, for perpendicular incident tracks. In the efficiency plots the hits reconstructed over the whole module are projected on a $2\times 2$ pixel cell window to put in evidence the sensor geometry and the possible effects of the columnar electrodes.

The geometrical inefficiency due to the columnar electrode diameter (5 $\mu{\rm m}$) was estimated to be around 1.5%. This effect can be greatly reduced by tilting the module on the beam. In non irradiated modules at 34∘ tilt angle, the hit efficiency detection reaches 99.3% for all sensors. The hit detection efficiencies as calculated in our data analysis for different runs are reported in Table 1.

In some selected runs it was possible to take data with a different bias voltage in order to verify if the depletion region reaches the $p^{+}$ electrodes delimiting the cell perimeter. The hit efficiency as a function of the applied bias voltage is reported in figure 6 for the available different bias points.

5 Conclusions

The upgrade of the CMS tracker for the HL-LHC will require radiation hard pixel sensors because of the expected extreme fluences. Thanks to their intrinsic characteristics, 3D pixels are to be considered as a possible option for the inner layers of the future trackers. Test beam results obtained with FBK 3D pixel sensors bump-bonded to the RD53A readout chip show high hit efficiencies before irradiation already at moderate bias voltages. After irradiation to 1\times$$10^{16} neq cm*-2* the efficiency is greater than 96.6% at bias voltage of 150V or less. A new batch of 3D pixels with stepper technology is in progress at FBK foundry since December 2018; it will provide very useful data for a deeper study of 3D sensors and their applications in large quantity. More irradiations and beam tests are planned to verify at larger scale the performance of these innovative detectors for their use in the upgrade trackers.

Acknowledgments

I wish to warmly thank Mirko Brianzi for his invaluable help in the construction and wire-bonding of the pixel modules. We thank the RD53 Collaboration for the RD53A chip; our results are not on chip performance but on sensor performance. This work was supported by the H2020 project AIDA-2020, GA no. 654168.