Semiconductor Tracker

From this point going radially inwards, the tracking volume is made up of semicon- ducting silicon sensors. These gossamer detecting sheets are used to provide radiation- resistant position measurements without adding too much material budget to the de- tector, which would reduce the accuracy of particle calorimetry that takes place at larger radii.

A semiconducting sensor works in a similar manner to the ionisation chambers men- tioned in earlier sections. A passing electrically-charged particle will excite atomic elec- trons in the medium it passes through to make a charge pair. One set of this pair (usu- ally the more mobile electrons) is accelerated to an electrode, amplified and turned into an electric signal. The difference here is that instead of collecting charge from the ionisation of gaseous atoms, the passing of a charged particle through or nearby the active silicon (whether it be a strip or a pixel) induces an excitation of electrons from the valence band to the conductive band of a solid lattice, and the electrodes that collect charge are segmented to provide more accurate position measurements [82, p9]. The semiconductor sensors on the ATLAS experiment are designed in a similar fashion to that seen in Figure 4.13.

Figure 4.13: Cross-section of a typical semiconductor module. Image from [83].

The problem with silicon is that the resistivity ρ (= RA_l) of pure silicon is low [82, p12]. This mean the power dissipated to the sensor when collecting charges is extremely high8_{. The electrical properties of the silicon lattice can be altered by inserting different}

elements (impurities) into the lattice. This is called doping.

A silicon lattice has four neighbours, and in pure silicon, an electron is shared among each one of them. Doping atoms that are in Group-3 of the periodic table (they have three valence electrons) would mean that only three bonds are formed with 3 nearest neighbours, leaving the introduced impurity with one fewer pairs. This leaves a ‘hole’ in the lattice that would attract electrons from neighbouring atoms causing this hole to appear to move. This type of doping is called n-type doping. Conversely p-type doping is when atoms that are in Group-5 are introduced into the lattice causing an additional free electron to move around9.

8_{Pure silicon has a resistivity of 10}4_{Ωcm. So with the some trial sensor dimensions (thickness 300µm}

and area = 2cm2_{) would get a resistance of 600Ω. Using a field of 30V to accelerate electrons will induce}

a current of 200mA and a power dissipation into the sensor of 12W!

9_{This ‘spare’ electron is still bound electrically to the atoms and does not have the energy to escape to}

Now that silicon with the desired conductivity can been created by alternating the con- centration of dopants in the silicon substrate, the issue becomes how one removes the residually-bound electrons and holes. By introducing the n-type substrate with p-type detecting strips embedded, and applying an electric field, the electrons from the n-type silicon will drift to the p-type and fill the holes. This is referred to as ‘depletion’. Upon full depletion, the silicon is said to be ‘active’. The field used to deplete the silicon is the same one used to accelerate charges that enter the conductance band [84, p21].

The external electric field across the sensor accelerates these charged particles to a col- lection point which in this case is a strip, creating a current. A silicon sensor in the SCT will typically have 768 strips. To read out the current and gain information about the position of the charge, each strip needs to be dealt with separately. Hence each strip is connected to an input channel on an readout chip called a ABCD3TA chip. This current is amplified in front-end electronics of the ABCD3TAs and turned into a binary readout signal, and transmitted into software logic which will store the component that read out the current, and the time window this occurred in. This is a ‘silicon hit’.

Since these ABCD3TA’s have 128 channels, six chips are needed to read out one sensor. As the strips are 12 cm long, the position resolution can be increased by placing two modules back to back with a relative rotation of 40mrad with respect to each other [85, p4]. Six ABCD3TA’s are mounted on a kapton circuit to form a hybrid. The hybrids are attached to two sensors via wire bonds to form a module, as shown in Figure 4.14. In the barrel of the SCT, 12 rectangular modules10_{are mounted onto a ladder to provide}

additional support, rigidity and common cooling and optical services [86, p51]. Going inwards from the outermost micro-strip sensor layers, the four concentric cylindrical layers in the barrel are formed from 56, 48, 40 and 32 full length ladders respectively, which leads to a total of 2112 modules (4224 sensors). In each endcap of the SCT, there

Figure 4.14: Picture of a SCT barrel module. The middle electronics board is placed over the junction of two sensors. Image from [69, p65].

are 9 rings of 52 trapezoidal-shaped modules for a total of 936 modules (1872 sensors).

One of the limiting factors in the efficiency of data collection on ATLAS is the archi- tecture of the read-out chips. This determines how fast data is read out and how long information can be stored before a decision on whether to keep or discard it has to be made. To cope with the increase of particle flux expected in the upgrade to the HL- LHC, the ABCD3TA chips are going to be replaced. Some of the work I did in this regard is discussed in Section 4.5.