Silicon detectors in High Energy Physics experiments

• Dr. Manfred Krammer, CERN, Geneva, Switzerland

 

Introduction and History

Silicon detectors are widely used in experiments in particle physics. The application of this detector technology is mostly for tracking detectors, i.e. detectors which measure the position of charged particles. From this information, track reconstruction software from the experiments deduces many parameters. These parameters include not only the flight path of particles, but also the momentum of the particle through the curvature in a magnetic field, the vertex of the interaction and possibly a decay vertex, called a secondary vertex, from particles with long lifetimes (e.g. a tau particle, or a hadron containing a b-quark). In some experiments silicon detectors are used as the active layers in sampling calorimeters. This field of application might become increasingly important with the necessity to have high grain calorimeters to be used by the so-called particle flow reconstruction algorithm.

The first ever silicon microstrip strip detector for particle physics, a surface barrier sensor, was tested in 1980 (Heijne, 1980) and the first silicon detectors using the planar technology and implanted strips were installed in the NA11 fixed target experiment at CERN in 1983 (Hyams, 1983). With the signals from this silicon microstrip detector the decay of particles containing a c-quark was tagged and subsequently the first measurement of the decay time and the mass of D-mesons was performed.

Figure 1: The mounting of silicon sensor modules on a half shell for the tracker of the CMS Experiment.

Following the rapid evolution of micro-electronics, which allowed the development of Application Specific Integrated Circuits (ASIC), and progress in the interconnection technologies (such as micro bonding), silicon strip detector systems were miniaturized and subsequently used by all collider experiments since. Silicon vertex detectors employed in all LEP experiments at CERN and in the SLD experiment at SLAC opened a new physics window through the reconstruction of b-quark decays. The fact that silicon detectors have a high inherent radiation resistance enabled the use of them in the harsh environment of hadron colliders. At the CDF and DØ experiments at Fermilab the silicon vertex detectors played a crucial role in the discovery of the top quark. At the LHC the CMS experiment made the next step by using silicon strip detectors not only in the vertex region but for the whole tracking system (CMS, 2008). In about 30 years silicon detectors used in high energy physics experiments have grown from a surface area of 24 cm² (NA11) to about 200 m² (CMS).

Figure 2: View of the silicon tracker of the ATLAS experiment.

Figure 1 shows a detail view of the CMS Inner Tracker and figure 2 the completed ATLAS Silicon Tracker.

Working Principle

The large majority of silicon detectors used presently are realized as so-called planar structures. This indicates the common production process in semiconductor industry, whereby in subsequent production steps the surface of a silicon wafer is structured with a photolithographic process and then treated by etching, implantation, material deposition, etc. Many of these “planar” process steps are followed after each other to achieve the required structures.

The bulk material is either an n- or p-doped silicon wafer, typically produced by the float zone (FZ) process which enables the production of silicon ingots with high specific resistivity (> 1 kΩ⋅cm). In the future material produced by the Czochralski (Cz) or Magnetic Czochralski (MCz) process could become of interest in experiments within an extreme radiation environment, due to the higher oxygen content which is beneficial for the radiation tolerance. In single-sided detectors, one side of the wafer is then structured (e.g. into strips or pixels) with implantation of the opposite charges. For example, p⁺ doping in n-type wafers and n⁺ doping in p-type wafers. These shallow implanted regions form a pn-junction with the bulk material – a diode structure. To create a deep depletion zone, void of free charge carriers, an external reverse bias voltage must be applied. The depth of this depletion zone in the bulk and in the implant region can be calculated as\[w\approx\sqrt{\frac{2\epsilon|V|}{e|N_{eff}|}}\] \(V\)…external voltage, \(N_{eff}\)…effective doping concentration, \(\epsilon\)...electric permittivity of silicon, \(e\)…elementary charge

Silicon usually contains both n-type and p-type impurities, hence \(N_{eff}\) is defined as \(N_{eff}=N_{a}-N_{d}\) with \(N_{a}\) and \(N_{d}\) the concentrations of acceptors and donors respectively.

Figure 3: Full depletion voltage versus resistivity for two different sensor thicknesses (300 µm and 500 µm).

Due to the fact that the effective doping concentration is lower in the bulk, compared to the implanted structures, the depletion zone develops deeper into the bulk and, eventually, if the reverse bias voltage is high enough, reaches the backside of the wafer. The voltage at this point is called the full depletion voltage \(V_{fd}\) which can be approximately written as\[V_{fd}\sim\frac{d^2}{2\epsilon\mu\rho}\] \(d\)…detector thickness, \(\mu\)…mobility of electrons (holes) in n-type (p-type) bulk, \(\rho\)…specific resistivity of the bulk

For a typical n-type detector with a thickness of 300 µm and a specific resistivity of 1 kΩ⋅cm the depletion voltage is approximately 300 V (see also figure 3).

At the depletion voltage or above, the full volume of the detector is sensitive to charged particles, which create electron-hole pairs along their path through the bulk. The electric field created by the reverse bias voltage separates the electron-hole pairs before they recombine again, and the electrons (holes) drift through the bulk to the positive (negative) voltage connection. The drift of these charges correspond to a current pulse, which is measured by external electronics.

Strip Detectors

In silicon strip detectors the implants are thin strips (typically around 20 µm wide with an interstrip distance of 50 -100 µm). A sketch of such a detector is shown in figure 4. In this example a sensor with n-doped bulk is explained. The cut through the structure shows the p⁺-implants on top, covered by single or multiple layers of an insulator (SiO₂ and usually a second layer of Si₃N₄), and an aluminum layer on top. The input of an amplifier is connected to the aluminum pad. The insulation layers form an integrated capacitor, which blocks the leakage current of the strips, but allows the high frequency signal to pass. This structure is called an AC coupling. The backside of the detector is heavily n⁺ doped and also covered by an aluminum layer. The doped region together with the aluminum layer allow for a proper ohmic contact to the backside.

Figure 4: Cut through of an n-type AC coupled strip detector.

Figure 5 shows a cut through of a detector along the strips. AC coupled strip detectors need biasing structures to apply the bias voltage to the strip implants. The most frequently used method is polysilicon resistors between each strip implant and a common bias line. The probe pad is an aluminum contact to the strip implant and allows the measurement of strip parameters such as the single strip leakage current. During operation this contact is not used. The guard ring structure shields the active sensor area from the cut edge region. Figure 6 is a picture from a strip detector taken with a microscope.

Figure 5: Cut through of an n-type AC coupled strip sensor along the strip implant showing the biasing structure.

The distance between the strip implants is the dominant parameter determining the position resolution of the detector. Using readout electronics, which allow measurement of the pulse height of the individual strip signals an interpolation between strips is possible and a precision of a few micrometer can be easily achieved depending on the signal-to-noise ratio. The upper limit of the position resolution in the case of a binary readout or low signal-to-noise ratio is the digital resolution given by \(\frac{strip~pitch}{\sqrt {12}}\).

Figure 6: A detailed view of a strip sensor. Visible on the left side are the polysilicon bias resistors, the probe pads, and the bond pads (from left).

Variants of this AC coupled strip detector are double-sided strip structures. In a double sided sensor the backplane is also structured, for example into n⁺ strips (in an n-doped wafer), which are orientated with an angle with respect to the strips on the junction side. However, positive oxide charges in the backside oxide generate an accumulation of electrons underneath the oxide, which electrically shortens the strips. For a functioning detector this accumulation layer has to be interrupted. Figure 7 shows a drawing of a strip sensor which uses p⁺ implants, p-stops between the n⁺ strips to interrupt the accumulation layer. Another option is the p-spray technique, where the whole surface of the sensor is implanted with a low density of acceptor atoms (p⁺).

Figure 7: Interruption of the electron accumulation layer with p-stops between the n⁺ strips.

With both sides of the detector instrumented, a two dimensional space point of the particles flight path can be reconstructed using a minimal amount of sensor material. This more complex and therefore also significantly more expensive detector type is only used in experiments were the material budget is of outmost importance, for example in the vertex region of precision experiments at e⁺e^- colliders (BelleII, 2010).

Hybrid Pixel Detectors

In parallel to increasingly sophisticated strip detectors hybrid pixel detectors were developed. The implants in these detectors are small pixels rather than strips, with dimensions such as 100x150 μm². The difficulty of these devices lies in the connection of the large number of pixels to the individual electronic channels. Each pixel of the sensor, realized on high-resistivity silicon material, has to be electrically connected to the corresponding input channel of the electronics chip. In hybrid pixel detectors these connections are done by a process called bump bonding.

Figure 8: Schematics showing a cell of a hybrid pixel detector.

A schematic drawing of a cell of a hybrid pixel detector is shown in figure 8. In this drawing the silicon pixel sensor is at the bottom. Small conductive bump balls (using materials such as In or SnPb) connect the pixels to the input pads of the electronic chip at the top. The layout of the electronic chip has to match the pattern of the pixel sensor. Large hybrid pixel detector systems are employed by the ATLAS and CMS experiments (ATLAS, 2008), (CMS, 2008). The achievable position resolution depends, as in the case of strip detectors, on the pixel dimensions and on the electronics. If the signal height is measured and used to interpolate between the pixels a position resolution of a few micrometers is achievable. The strength of hybrid pixel detectors is their capability to operate in very high track density environments making them the ideal choice for the detectors closest to the interaction point.

Other Sensor Structures

Apart from strip and hybrid pixel detectors, which make the majority of detectors used in high energy physics experiments a large number of variants have been developed. Driven by the fact that hybrid pixel detectors are complex assemblies with large number of connections and also require a significant amount of material, several groups have developed monolithic pixel detectors. These detectors combine the sensor volume and the first electronics stage within one device. In the following example some of these structures will be introduced with references to more details.

Charged Coupled Devices (CCD) have been in use for a long time in high energy physics experiments and in astronomy. CCDs have shallow depletion regions and collect electrons by diffusion. These electrons are then shifted through the columns and through a final row to a single readout channel. Consequently, the readout is slow and the device may be only used where speed is not critical (Abe, 1997).

Other monolithic detector structures are:

Monolithic Active Pixel Sensors (MAPS) use an n-channel MOSFET transistors (NMOS) embedded in an epitaxial p-layer similar to standard CMOS chips. The n-well of the transistor collects the electrons generated by charged particles from a thin depletion layer through diffusion only (Turchetta, 2001). A deeper depletion layer and hence larger signals can be achieved by applying higher bias voltages or by the use of high resistivity epitaxial layers. In these structures, named HV-CMOS and HR-CMOS, the electrons are collected through drift (Peric, 2007).

The Silicon on Insulator (SoI) sensors overcome the problem of the low signal from only partially depleted sensors by combining a high resistivity silicon sensor wafer with a low resistivity electronics wafer, being chemically bonded together. The transistors implemented in the electronics wafer are connected to the implant of the sensors, which is fully depleted (Arai, 2010; Marczewski, 2005).

A Silicon Drift Detector (SDD) has parallel p⁺ strips on both sides of the high resistivity n doped wafer. Created electrons accumulate in a potential valley within the bulk and are transferred during readout to the anodes. The two dimensional measurement comes from the strip signal and from the measurement of the drift time of the electrons to the anode (2^nd coordinate) (Rehak, 1986).

Also in DEPFET (Depleted P-channel Field Effect Transistor) detectors a potential valley for electrons is created underneath the p⁺ strips within the n bulk. The accumulated electrons however drift underneath the gate of a field effect transistor and modify the source drain current. The DEPFET detector has therefore a built-in amplification. Following the readout the accumulated electrons have to be swept away by an active clear (Kemmer, Lutz, 1987).

Another development are 3D detectors (Parker, Kenney, Segal, 1997). These detectors employ a different concept, in the sense that the depletion does not develop from horizontal implants, but from vertical columns etched into the bulk. These columns are filled with n⁺ and p⁺ doped material. The advantage of 3D detectors is the small distance between the columns and hence the smaller depletion voltage needed for operation. Consequently lower voltages are needed for highly irradiated sensors, which leads to higher radiation tolerance – the main motivation to use 3D sensors.

Silicon Detectors under Radiation

Silicon detectors are very radiation tolerant and in many cases the only possible choice for detectors in areas of very high radiation, for example in the inner region of hadron collider experiments. This, however, does not mean that silicon is not damaged by radiation. Depending on the type of radiation one does observe the following effects.

Light particles, such as electrons, positrons and photons create permanent charges within the amorphous silicon on the surface of the sensors, which alters the potentials in the interface between the silicon dioxide and the silicon. In a typical sensor this leads for example to an increase of the interstrip capacitances. These particles, however, very rarely damage the monocrystalline structure in the bulk, due to the rather low energy transfer.

Heavy particles such as protons, neutrons and pions have enough energy to also dislocate silicon atoms from their places in the silicon lattice. These defects can be simple empty lattice locations (vacancies) or silicon atoms between lattice locations (interstitials), or more complex defects also in connection with impurity atoms present in the bulk. Very hard impacts can even produce cluster defects, involving and dislocating as many as 100 silicon atoms. All defects in common is the introduction of new energy levels within the band gap between the valence band and the conduction band of the semiconductor. These additional levels are the cause for the change of detector parameters, i.e. the increase of the leakage current, the modification of the effective doping concentration, and the creation of trapping centers reducing the lifetime of the charge carriers.

The increase of the leakage current can be well described by the formula \(\Delta I = \alpha\phi_{eq}\) with \(\alpha\) the damaging constant and \(\phi_{eq}\) the particle flux normalized to 1 MeV neutrons (a convenient definition to enable the comparison of the damage by different particles and energies). The measured increase of the leakage current depends on the ambient temperature during the irradiation and during the period before the measurement is taken. Annealing effects decrease the currents, until saturation is reached. \(\alpha\) derived from measurements is about \(\alpha\sim 4 \times 10^{17}\) A/cm (after 80 minutes annealing time at 60\(^{\circ}\)C) (Moll, Fretwurst, Lindström, 1999). This number is valid for all standard structures and silicon materials tested.

Figure 9: Dependence of \(V_{fd}\) (left axis) and \(|N_{eff}|\) (right axis) on the radiation fluence.

The dependence of the effective doping concentration \(N_{eff}\) and the full depletion voltage on the radiation fluence is shown in figure 9. Due to the fact that radiation dominantly creates acceptor like defects and only to a smaller extent donors, the silicon becomes increasingly p-doped. As a consequence n-type silicon sensors undergo a type inversion and become quasi p-type sensors. In n-type detectors after type inversion the pn-junction is moved to the backside formed by the n⁺ backside implantation and the now p-bulk. As long as a bias voltage can be applied as required by the increasing depletion voltage and the sensor remains fully depleted it remains also essentially functioning. However, due to a longtime effect called reverse annealing, the effective doping concentration continues to increase even after irradiation. The consequence for the experiments is, that once the sensors have been irradiated they have to remain at low temperature even in periods where the experiment is not operated to prevent this effect from progressing.

In the present LHC experiments the detectors are designed to withstand the radiation of about 10 years of operation and an accumulated fluence of about 10¹⁴ particles/cm². For the upgrade of the tracking detectors of the LHC experiments it is envisaged to use sensors on p-type material, thus avoiding type inversion. For future experiments with even higher radiation loads the development of silicon materials with greater oxygen content, which reduces the increase of the effective doping concentration, or the development of special structures, such as 3D detectors, is ongoing.

References

Heijne, E.H.M., et al. (1980). A silicon surface barrier microstrip detector designed for high energy physics. Nucl. Instr. and Meth. 178: 331-343.

Hyams, B., et al. (1983). A silicon counter telescope to study short-lived particles in high-energy hadronic interactions. Nucl. Instr. and Meth. in Phys. Res. 205: 99-105.

CMS (2008). The CMS Experiment at the CERN LHC. JINST 3 S08004

Belle II (2010). Technical Design Report. KEK Report 2010-1

ATLAS (2008). The ATLAS Experiment at the CERN Large Hadron Collider. JINST 3 S08003

Abe, K., et al. (1997). Design and performance of the SLD vertex detector: a 307 Mpixel tracking system. Nucl. Instr. And Meth. in Phys. Res. A 400: 287-343

Turchetta, R., et al. (2001). A monolithic active pixel sensor for charged particle tracking and imaging using standard VLSI CMOS technology. Nucl. Instr. and Meth. in Phys. Res. A 458: 677-689

Peric, I. (2007). A novel monolithic pixelated particle detector implemented in high-voltage CMOS technology. Nucl. Instr. and Meth. in Phys. Res. A 582: 876-885

Arai, Y., et al. (2010). Developments of SOI monolithic pixel detectors. Nucl. Instr. and Meth. in Phys. Res. A 623: 186-188

Marczewski, J., et al. (2005). Monolithic silicon pixel detectors in SOI technology. Nucl. Instr. and Meth. in Phys. Res. A 549: 112-116

Rehak, P., et al. (1986). Progress in semiconductor drift detectors. Nucl. Instr. and Meth. in Phys. Res. A 248: 367-378

Kemmer, J., Lutz, G. (1987). New detector concepts. Nucl. Instr. and Meth. in Phys. Res. A 253: 365-377

Parker, S.I., Kenney, C.J., Segal, J. (1997). 3D — A proposed new architecture for solid-state radiation detectors. Nucl. Instr. and Meth. in Phys. Res. A 395: 328-343

Moll, M., Fretwurst, E., Lindström, G. (1999). Leakage current of hadron irradiated silicon detectors – material dependence. Nucl. Instr. and Meth. in Phys. Res. A 426: 87-93

Sponsored by: Laurent Serin, LAL - IN2P3 - CNRS, Orsay, France

Reviewed by: Dr. Michael Moll, CERN, Geneva, switzerland

Reviewed by: Dr. Clara Nellist, LAL/CNRS, Orsay, France

Accepted on: 2015-10-01 08:52:34 GMT