Study of prototypes of LFoundry active and monolithic CMOS pixels sensors for the ATLAS detector

High Energy Particle Physics experiments at the LHC use hybrid silicon detectors, in both pixel and strip geometry, for their inner trackers. These detectors have proven to be very reliable and performant. Nevertheless, there is great interest in the development of depleted CMOS silicon detectors, which could achieve similar performances at lower cost of production and complexity. We present recent developments of this technology in the framework of the ATLAS CMOS demonstrator project. In particular, studies of two active sensors from LFoundry, CCPD_LF and LFCPIX, and the first fully monolithic prototype MONOPIX will be shown.


Towards HL-LHC
The Large Hadron Collider, is currently taking data at 13T eV and at a luminosity of 10 34 cm −2 s −1 . It is expected that the center of mass energy of the machine will increase to 14T ev for run 3. Then, the machine is scheduled to undergo an upgrade, denoted as the High Luminosity LHC or HL-LHC, that will increase the instantaneous luminosity to about 7 · 10 34 cm −2 s −1 . In this phase, the bunch crossing time will remain 25ns , while the number of proton collision per bunch crossing will be increased. This luminosity increase will set new requirements to the detectors in terms of speed, data acquisition rate and radiation resistance. In particular, the tracking system, the innermost part of the experiment, will be replaced with an all Silicon tracker, the ITK. This will be divided in two parts, the inner tracker, made of Silicon pixel detectors, and the outer tracker, made of Silicon strip detectors. The devices presented in this paper are potential candidates for the 5th (last) layer of the pixel system, at about 30 cm from the interaction vertex (see Figure 1). At that distance the fluence and the Total Ionization Dose (TID) are expected to be 1.5 · 10 15 n eq /cm 2 and 80M rad respectively.

Silicon Hybrid detectors
Most current HEP experiments at the LHC have a pixel system based on Silicon Hybrid technologies. These are produced by coupling a high resistivity Silicon diode with a front end read-out chip. The connection is guaranteed by arrays of conductive bumps, one per pixel, generally made with Indium. Picture 2 shows a sketch of this process. These detectors have achieved excellent performances providing good signal response within 25ns. They are also radiation hard up to 5 · 10 15 n eq /cm 2 . Nevertheless, this technology, although very well understood, remains very difficult and expensive to produce, since the bump bonding process is slow, complicated and with high risk of failure. For this reason, there is a lot of interest for novel detectors called depleted Monolithic Active Pixel Sensors (DMAPS).

CMOS detectors: from standard applications to HEP
The CMOS technology provides a different concept of detector, in which the read-out circuitry is inserted inside the Silicon sensor itself (for this reason they are also called MAPS). This allows an easy production at lower cost, with less material. These sensors has been used in a variety of applications from industry to Medical Physics (see Figure 3(a)). Standard MAPS detectors are not suitable for LHC and HL-LHC operation, since in these detectors the charge collection happens trough thermal diffusion, which leads to a charge collection time that is too long for these challenging environments. For this reason, some improvements must be implemented in the technology. Specifically, the resistivity of the substrate and the bias voltage must be increased, since the depletion region and therefore the charge collected is proportional to √ ρ · V . The read-out circuitry should be nested inside a deep N-well, providing isolation to operate at high voltages. This way, the depletion region is formed between the substrate and the deep N-well itself [1]. A sketch of these add-ons can be seen in Figure 3(b).

The ATLAS CMOS pixel collaboration
The ATLAS demonstrator program aims at proving the feasibility of the DMAPS technology for the ITk. Different foundries and strategies have been investigated. One of the main feature to study is the fill factor, i.e. the ratio between the surface of the deep N-well, shown in Figure 3(b), and the total surface of the pixel. In particular, with a large fill factor the read-out is inside the collection well, while with the small fill factor it is outside. The first option can achieve higher bias voltages and signals but is affected by higher noise. The second option is optimized to minimize the capacitance and therefore the noise and the power dissipated. The small fill factor approach has been developed in the design and production of prototypes with TowerJazz foundry, while the large fill factor has been implemented with LFoundry and ams. Herein we describe the current status of the studies performed to evaluate the LFoundry process.

LFoundry prototypes
In the past three years, three prototypes have been designed and fabricated in the LFoundry technology: CCPD LF, LF-CPIX and LF-MONOPIX.

Active CMOS devices
CCPD LF and LF-CPIX have a very similar design: in both cases amplifiers, discriminators and other signal processing are implemented insides a deep Nwell. For this reason they are referred to as active CMOS devices. In addition, both have different kinds of pre-amplifier: PMOS, NMOS and CMOS (the last only for LF-CPIX) [2]. The basic read-out scheme is sketched in Figure 4.
The main difference between the two devices is the pixel size, which is 33 × 125µm 2 for CCPD LF and 50 × 250µm 2 for LF-CPIX, the latter being the  standard ATLAS pixel size (same as FE-I4). The layout of the two active pixel prototypes can be seen in Figures 5(a) and 5(b). Looking at the front layout, one can easily notice the large fill factor as the deep N-well occupies most of the pixel surface. Figure 5(b), in particular, show also the lateral layout, where the multiple nested wells can be seen. Both devices can be either capacitively glued to an FE-I4 chip or read-out through a standalone board.

Fully Monolithic device
The last prototype, LF-MONOPIX, is a fully monolithic device, i.e. it implements the whole read-out inside the pixel itself. As shown in Figure 6, the digital part of the read-out, which follows the FE-I3 architecture, was added on the right-hand side of the analogue pixel based on LF-CPIX. LF-MONOPIX delivers a serialized output with a 40MHz clock (25ns resolution) and no read-

Characterization and results
CCPD LF, LF-CPIX and LF-MONOPIX have been extensively characterized within the collaboration with different tests and irradiation doses and sources. Some results are shown in this section, while others are currently being collected and processed.

CCPD LF performances before irradiation
CCPD LF characterization with MIPs was performed at ELSA facility in Bonn with a 3.2GeV electron beam. The collected spectra show a distribution compatible with a Landau (see Figure 7(a)), and the dependence of the Maximum Probability Value (MPV) parameter on the bias is compatible with a resistivity of 3kΩ · cm (see Figure 7(b)). This result has been confirmed by investigating the properties of charge collection with edge-TCT, a technique that consists in shining very fast Infra-Red laser pulses on the edge of the sensor[3]. Once the laser is well focused, one can study the signal produced at different depth levels inside the bulk and have a direct measurement of the depletion depth. Figure 8(a) shows the charge profiles obtained with this method at different bias voltages, showing that the depletion region and therefore the signal increases with bias voltage as expected. The depletion depth is then calculated as the FWHM of this profile, and its value is plotted as a function of the bias voltage in Figure 8(b). From an interpolation with a square root function it was possible to estimate the resistivity to be 3.3 ± 0.5kΩ · cm, which confirms the result obtained with an electron beam.

LF-CPIX performances before irradiation
In the same way as CCPD LF, the properties of charge collection for LF-CPIX have been investigated with edge-TCT. Figure 9(a) shows the charge collection profile as a function of the laser depth for different bias voltages. From an interpolation with a square root function (Figure 9(b)) it was possible to estimate the resistivity to be 4.2 ± 0.3kΩ · cm.

CCPD LF performances after neutron irradiation
Some CCPD LF prototypes have been irradiated with neutrons in Ljubljana up to a fluence of 5 · 10 15 n eq cm −2 and then tested with the edge-TCT to evaluate the reduction in the depletion region due to bulk damage. The results are shown in Figure 10. From this plot, it is clear that even after a fluence of 2·10 15 n eq cm −2 a depletion depth of more than 50µm can be obtained, surpassing the requirements for operation in the outer pixel layers.

LF-CPIX performances after proton irradiation
Some LF-CPIX samples have been irradiated with 27 M eV protons to a fluence of 10 15 n eq cm −2 at the Birmingham Cyclotron[5]. Afterwards, edge-TCT has been performed to check the effect of the irradiation on the charge collection. The results, shown in Figures 11(a) and 11(b), prove that the device can be depleted to a depth of more than 70 µm after this dose. In addition, an X-ray characterization with 55 F e proves a good behaviour of the device (Figure 12), as the expected peak at 1640e − has been observed.

Conclusion
The ATLAS upgrade is going to impose very high standards on the performance of its new tracking system. A new concept of Silicon detectors, based on the CMOS technology, is under investigation as a candidate for the fifth pixel layer.
The new technology brings great advantages and allow easy and low cost of production of detector modules. The ATLAS demonstrator program is studying different prototypes and foundries to prove the feasibility of this technology for the ATLAS upgrade. One of the foundries that have been investigated is LFoundry, which provides sensors with high resistivity and high voltage additions. These add-ons ensure a sufficiently high signal. Three devices have been designed and produced with this foundry, two active CMOS sensors, CCPD LF and LF-CPIX, and a fully monolithic device, LF-MONOPIX. Many tests have been performed on these devices before and after irradiation with with neutrons, protons and X-rays The tests performed so far indicate that these devices are fully functional at the needed high radiation doses. The next step for the collaboration is then to produce a full CMOS module to finally prove the feasibility of this technology for the ATLAS upgrade.