While the new Inner Tracking System detector, ITS2, could finally start its data taking in nominal operation for the RUN3 of LHC, ALICE is already working at replacing the three innermost layers during LS3. The ITS3 (Fig.1) is a new vertex detector, that will significantly increase the performance in spatial resolution, improving the precision of several measurements in the heavy-flavour sector, with important extensions at low p_T and bringing another set of fundamental observables into reach.

Figure 1: Design of the new ALICE Inner Tracking System 3 (ITS3). Exploded CAD model showing the 3 half-layers connected by carbon foam rings and longerons to an external cylindrical structure that provides the main support.

The design of the new vertex detector aims to reduce the material budget of the first detection layer to an unprecedented minimum of 0.05% X0, and to get closer to the interaction point at a radial distance of 18 mm. Achieving such a challenging objective on the material budget, essentially requires having in the active area only the thin MAPS silicon sensor (<50μm), while removing mechanical structures, cooling lines and any printed circuits for the interconnection of adjacent sensors.

The solution is as simple as designing a large silicon sensor, cooled by an airflow, and bent to a cylindrical shape that covers the entire layer and that requires electrical connectivity only at the edge Fig 2.

Figure 2: 50-micron thin silicon sensor, 280mm in length, bent to 30mm radius. Stitching process used to obtain large area MAP sensors. Electrical connectivity at one edge of the sensor by wire bonding. Carbon foam provides mechanical and thermal functions.

Is this feasible?

ALICE has started this assessment by exploiting the mechanical properties of thin silicon chips (15mmx30mm), the ALPIDE, and demonstrating that they are flexible enough to be bent onto truly cylindrical surfaces of radii well below 18mm. Their performance was validated and compared to unbent sensors.

In parallel, ALICE chip designers were dealing with stitching technology to overlap sensors’ images on a large silicon wafer, such to cover the surface of an entire ITS half layer (280mmx93mm); this removes the need of printed circuit boards for interconnection in the active area, with signals and power feeding the sensors only at its edge.

These developments show a promising path to a large silicon layer, bent to cylindrical shape, that will not only reduce the achievable distance from the beam pipe (no staggering of staves will be needed), but that will also give the layers an intrinsic stability, minimizing the requirement on supporting structure.

Carbon foam, with large radiation length, has been identified as the design choice for the mechanics of the new vertex detector to achieve the requirement on minimum material budget. The carbon foam, largely used in the present HEP tracking detectors as thermal filler between the sensors and the cooling metallic pipe, is intended, instead, to realize the ITS3 sensors support and act as a radiator for the air cooling. Indeed, the low power consumption in the pixel matrix (~20mW/cm²) that can be achieved with MAPS, and the operational ambient temperature, suggest adopting an air-cooling solution while removing liquid cooling and pipes from the active area.

For the development of the mechanical and cooling design, the starting point has been to demonstrate the feasibility of bending such a large and thin silicon layer, and subsequently to produce bent layers for the prototyping and testing phases. Three different silicon sizes were considered, each size corresponding to one of the three half-layers into which the detector is split, so as to be mounted around the beampipe (radius L0= 18 mm, L1=24 mm, L2=30 mm). Bending these layers to the correct radii has required several trials and the development of devoted jigs and procedures to be performed exclusively in a clean room environment (Fig.3).

The flat wafer-size sensor is rolled against a dedicated mandrel which provides the nominal cylindrical shape. An FPC at each side is aligned and bent to match the edges of the sensor, where the electrical interconnection is achieved by wire bonding. The possibility to reliably perform an effective wire bonding on curved interfaces has been demonstrated.

Figure 3: Naked Silicon 50 um thin, bent on its mandrel with carbon foam glued in position.The three half-layers assembly sequence.

Once the electrical connection step is completed the FPCs and the half-layer sensor are ready for the mechanics bonding. While still sitting on the cylindrical mandrel, carbon foam structural components are glued at the periphery of the sensors through an optimized procedure.

Figure 4: Exploded view of half barrel assembly (Engineering model) and the air-cooling-ducts for the 3 layers. At the periphery of the sensors, where most of the power is dissipated, the thermal performance is enhanced by the use of carbon foam that act as a radiator.

At this stage, the outermost half-layer (h-L2) is glued, through the carbon foam components, to a carbon composite cylinder that acts as an exoskeleton. This allows the removal of the mandrel, the cylindrical shape now being guaranteed by the external exoskeleton.

In a similar way half-layer1 is glued to half-layer2 and half-layer 0 to h-layer1 in sequential steps such to realize the entire half barrel, only supported by the carbon exoskeleton shell.

The carbon foam components glued on the half-layer sensor consists of two carbon half-rings and two longerons along the sensor sides (Fig4). They all fulfill a structural function, holding in position the half-layer sensor, while one of the two half-rings also acts as a radiator for the air cooling, being glued at the sensor edge (~5mm wide) where most of the power is dissipated (<1W/cm²). This carbon foam half-ring largely enhances the thermal performance by increasing the surface of exchange with the air through its open cell geometry.

The choice of the carbon foam has been then tailored to achieve minimum material budget for all the structural components while, for the half-rings acting as a radiator, the thermal conductivity has been favored. This has resulted in the use of two different carbon foams, Duocel® manufactured by ERG Aerospace for structural (45kg/m3, k=0.05Wm/K), and K9 foam manufactured by Lockheed Martin for thermal (200 kg/m3, k=25Wm/K).

The mechanical and thermal characterization of the foam, the effect of glue penetration, the correct representation and simulation of the foam in a CFD analysis, and the final design validation through a wind tunnel test, have required a large R&D effort shared between ALICE and the EP R&D program on technologies for future HEP detectors.

The mechanical design concept based on the exclusive use of carbon foam as mechanical support has been demonstrated to work properly (Fig.4).

Figure 5 Carbon foam half-ring. It enhances cooling performances at the edge of the sensor (bottom left) where most of the heat is dissipated and controls the flow distribution (bottom right).

The interface between carbon foam and silicon has been identified as one of the key design features, affecting both thermal and mechanical performance. A special procedure, based on the use of a veil of pre-impregnated carbon fleece, interposed at this interface, has allowed for the control of the gluing penetration in the carbon foam, tuned to reach the best compromise between minimum material budget and thermal conductivity.

The properties of carbon foam as a radiator have been fully exploited, in term of thermal exchange and air flow impedance; the obtained parameters have been used in the CFD analysis and the carbon foam thickness and holes distribution improved to optimize the performance.

An integrated design of the entire half-barrel, constituted by three half-layers, has been developed and several engineering models built. The full characterization has required the design and construction of a wind tunnel setup. A series of tests conducted on the different half barrel models have allowed for a system design optimization. This has resulted in specific project choices like the dedicated 3D printed manifolds feeding each layer independently, and with an average air speed of 8m/sec. The most recent test results show that the air-cooling system can effectively remove the heat, both at the sensors periphery, where most of it is dissipated and within the sensor pixel matrix. An overall temperature uniformity is achieved within less then 5 degrees and the sensor can operates at a temperature about 8 degrees above the inlet air cooling temperature (20°C).

In parallel with the thermal performance assessment, the wind tunnel tests have been used to carefully monitor the stability of the large half-layer under the airflow. Laser optical measurements of the detector surface have shown vibration below 1 µm i.e. well within requirements.

Now while working at the preparation of the TDR, the mechanical effort is addressed into the next challenge that is to demonstrate the ability to install and match the two half barrels, based on such a delicate sensors assembly, around the beampipe with sub millimeters precision. All this must be achieved under the additional constraint of a limited access, due to the Experiment layout, that allows only for a remote control 4 meters away from the interaction point where the matching is happening.