Review of Semiconductor Ionizing Radiation Detectors*

(School of Electrical Engineering, Faculty of Engineering, Tel Aviv University, Tel Aviv 69788 Israel)

Abstract：There are three main reasons for the importance of semiconductors in the detector field: (i) the usefulness of the penetrating radiation; (ii) the hazards of the radiation to living creatures; (iii) the information it carries about the nature surrounding us. There are several types of “detections”: (1) counting the particles; (2) finding the energy of each particle; (3) finding the time of the particle-matter interaction; (4) finding the total energy carried by the radiation per unit time, per unit area. There are various types of ionizing particles and various energy ranges. These particles have varying reaction/interaction cross-sections with matter; therefore, their detection requires various detectors.This presentation reviews the various semiconductor detector types (materials and structures), applications, and challenges.

Key words：semiconductor detectors; penetrating radiation; particle interaction with matter

1 Introduction and background

The ionizing particles are agents capable of removing a bonded electron from an atom (or a molecule), thus “ionizing” it. Such agents include fast moving ions (e.g., alpha particles), fast moving protons and neutrons, fast electrons (beta particles), fractions of protons and neutrons (e.g., quarks, muons, pions, etc.), as well as electromagnetic waves in the high end of the energy spectrum (mid UV and above).The German scientist Wilhelm R?ntgen is usually accredited with the discovery of the first ionizing radiation agent, the “x-ray”. He also gave that name to the agent, as a temporary “mathematical” notation (X represented unknown). Wilhelm R?ntgen used Crookes tubes to accelerate electrons in his research, and he published the first x-ray image of his wife’s hand in 1895 (using barium platinocyanide screen)[1]. He was the first to conduct and publish a systematic study, but he was certainly not the first scientist to observe and report effects of x-rays. In fact, the earliest observation of x-ray was reported to the Royal Society of London, more than a hundred years prior, in 1785 by a famous British scientist William Morgan. He passed an electrical current through partially evacuated glass tube, which resulted in glowing. Another important and preceding report of x-rays generated in vacuum tubes was by Fernando Sanford, in 1894 in Physical Review (describing his discovery as “electric photography”). The impact of the new invisible radiation on the scientific community was substantial and shortly after the man made source, the radioactive isotopes started appearing: uranium salt by Henri Becquerel in 1896, polonium and iridium by Marie and Pierre Curie in 1898. Although it was not immediately correlated to the radiation coming from the vacuum tubes. The hazards as well as the usefulness associated to the mysterious radiation were noticed immediately. Elihu Thomson conducted experiments on himself deliberately exposing his finger to x-rays and systematically recording the development of radiation burn.In 1896, Dr. William Lofland Dudley of Vanderbilt University reported local hair loss after he exposed his head to X-ray for 1 hour. It should be noted that Dr. Dudley experimented on himself in order to help a child who had been shot in the head. In 1920’s the case of “radium girls” shocked the United States, after workers became fatally ill from contact with “harmless self-luminous paint”. Severe regulations were imposed on working with radioactive materials after that case.The period between 1950’s and 1970’s brought up what was called the “particle zoo”, when new particles were found one after another forming the “standard model” which is widely considered as the foundation of the modern particle physics.In order to be detected, the ionizing particles (waves) must interact (or react) with the detector media. Such interaction transfers part of the energy to the detector media causing it to change its properties (e.g., film), temperature (e.g., colorimetry), change electron energy states followed by light emission (scintillators), or generate electrical charge (gas, and semiconductor detectors). Reaction of the radiation with the media generates new isotopes. This review focuses on semiconductor detectors with direct conversion, namely, the incident agents generate free charge carriers in the detector media (ionization).

2 Semiconductor detectors-materials applications and configurations

This section summarizes the basic modes of detection, applications and the compatibility of the common detector materials and configuration to these applications.

2.1 Modes of operation

Due to the above mentioned variety of particles and energies, there is not one particular optimal detector material/configuration. In addition, one has to consider the various common detection modes:

(1) Particle counting：the system counts the particles regardless of their energy (in a given range). Sometimes selectivity is required, namely focusing on a certain particle type. Thus, the figures of merit are selectivity, sensitivity (namely the lowest detectable energy), detection energy range of the particles, absorption efficiency, and the counting dynamic range.

(2) Energy spectroscopy：the system detects a single incident particle and provides the information about the particle’s kinetic energy. To obtain such information,the detector should absorb all of the particle’s kinetic energy and produce a corresponding electrical signal. The performance is described by the energy resolution, absorption efficiency (as a function of energy), selectivity, detectable energy range, speed of response, linearity, and maximum throughput rate.

(3) Total energy flux per unit area (flux mode)：in this mode the system does not discriminate between separate particles, but rather it provides information on the incoming energy flux (total energy of all the incident particles per unit time). Such system by definition integrates the incoming particle flux. The figures of merit are absorption efficiency, dynamic range, sensitivity (lowest energy of the dynamic range), speed of response, selectivity, linearity

(4) Time spectroscopy：the system provides information about the exact time of interaction, regardless of the particle energy transfer. The main figures of merit are the quantum efficiency, timing resolution, sensitivity, and counting rate.

Regardless of the mode of operation, the semiconductor detectors often exhibit afterglow and polarization. Both effects can be traced to deep levels that can be produced in course of material preparation or as a result of radiation induced damage. In energy spectroscopy and flux systems the electronic circuits collect the charge from the detector for a given time window before the information is considered valid and read out. This is done to distinguish between the events and to minimize the impact of noise and dc biasing currents. Afterglow is the gradual de-trapping of the signal charge that was trapped during charge collection after the collection time is over. In flux systems it hinders the intensity of the following effect (e.g., next frame in a scanning system). Polarization usually refers to the variations of detector performance with time. There are two types of polarization: related to dc biasing and to detected events. The former is in fact a turn-on transient (that could last for hours in some cases) while the electric field inside the detector reaches the steady state conditions. In this case if similar events occur over that time the readouts will be systematically different. The other polarization is induced not by biasing, but by the event generated charge. Part of the generated charge is trapped and modifies the electric field profile inside the detector. This leads to performance variations based on the history of events.

There are also general properties that are not related to the operation mode, but rather to application environment: radiation hardness and solar blindness.

2.2 Applications

Applications for ionizing radiation started developing almost as soon as it was discovered. The main fields are medicine (both, diagnostics and therapy), security/military, environment monitoring, fundamental science, energy system control, and nondestructive testing.

The first and most common application is for medical imaging, where the semiconductor detectors can replace the radiographic film, and furthermore, enable dynamic scanning. In some imaging systems the radiation source is external to the imaged item (e.g.,x-ray tube in computerized tomography systems, or a neutron source in neutron imaging systems); in other systems the source is internal (e.g., injected isotopes in nuclear cameras). In both cases，part of the radiation has to penetrate through the scanned item, to reach the detectors. This puts constrains on the particle type and energy. The first and still most common medical imaging is done by x-ray，where for good image contrast， photon energy should be tailored for the tissue type and thickness. Typically dental imaging is performed with bremsstrahlung (braking) peak at 30～40 keV; mammography optimum for contrast-dose is around 20 keV; computerized tomography systems use tubes peaking at ～100 keV, but substantially emitting energies even above 300 keV. In nuclear cameras radiopharmaceuticals are injected and their segregations are monitored. Typically, for different regions of interest in the body such radiopharmaceuticals include99mTc,133Xe,75Se (emitted gamma energies 80～400 keV). Whatever the type and the energy range of the particles, the absorption efficiency of the particles leaving the body is a critical parameter; in most medical systems close to 100% absorption is required in order to minimize the patient irradiation dose and to improve image quality.

With all that being said some x-ray imaging systems are intended for small objects (e.g., small animals, insects, ext.), in which case the photon energy is considerably lower.

Energy spectroscopy is the most demanding operation mode for the detectors and the electronics in terms of signal strength, noise, etc. In medical imaging, it is sometimes used to distinguish the original particle from scattered particles, which is critical for example in nuclear cameras. It is also critical for material identification (nuclear fuel analysis, environmental tests, x-ray photoluminescence, etc.). Often the energy of individual particles (with or without time correlation) is used to identify a physical/chemical process (e.g., plasma). For proper spectroscopy operation the photons should not only interact with the active detector volume, but rather they should transfer all their energy to it. Escape of secondary particles (e.g., scattered and characteristic photons) would hinder the operation. Therefore, typically low atomic number materials are incompatible. However, for some spectroscopic applications, the energy range of interest is between 1 and 10 keV. This is the energy range of the K-shell emission lines of elements with atomic numbers between 11 (natrum) to 30 (zinc). For such applications low atomic number elements could be fine.Particle detection and tracking systems are used for many years by scientific community in high energy physics. In tracking systems the particle must leave its signature in several detectors so that its track can be reconstructed. Therefore, each particle should be detected, yet only a small fraction of its energy should be transferred to each detector. In the common case of minimum ionizing particles the deposited energy is weakly dependent on the media atoms, yet is linear with the thickness. Thus detectors should be just thick enough to produce a reliable signal, and not much thicker in order not to hinder the particle energy and to reduce secondary particle showers.

2.3 Device architectures

In semiconductor detectors，the generated electron-hole pairs should contribute to external current in order for the interaction to be recorded. The two current type at the contacts are the induction current due to varying electric field, and the conduction current by the carriers physically recombining at the contact. Only the volume which contributes to these currents can be considered “active”. Full and fast charge collection are most desirable, thus drift current should usually dominate over diffusion. The common architectures are described in Fig.1.

The simplest technology-wise structure is MSM in Fig.1(a), which requires only fabrication of Ohmic or (Ohmic-like) contacts, and features uniform electric field throughout the device (full active layer), and no shot noise; however, the resistivity of the detector semiconductor has to be very high to keep the leakage currents no higher than nano-Ampere range even at high bias voltages. The junction case exhibited in Fig.1 (b), includes the p-n and the Schottky junctions. The current is limited by the thermionic emission over the junction barrier, thus lower resistivity materials can be used; however, the electric field is not uniform and the active volume is defined approximately by the depletion layer (low effective doping is required for thick active layers); in addition, high electric field reaching throughout the device may cause punch-through, increasing the dc current dramatically. Another disadvantage of the p-n junction configuration is the high detector capacitance (inversely proportional to the depletion region thickness). Ideal capacitance as such does not contribute noise, however it contributes considerably to the overall system noise by impacting the system transfer function (in other words, it increases the throughput of the noise from other elements to the output). The P-I-N structure, shown in Fig.1 (c) is widely used when both, p- and n-doping are available technologies. The lightly doped (ideally intrinsic) layer is depleted at relatively low voltage (called the full-depletion voltage), yet there is no punch-through due to the highly doped region beyond the depletion layer. Sometimes the P-I-N structure is enhanced by a multiplication region, as shown in Fig.1(d). In p-n junctions as well as P-I-N devices, the highly doped region has a week electric field, thus charge collection from that layer is poor, creating so called “dead-layer”. Some architectures like those shown in Fig.1(e) and Fig.1(f) require mature fabrication technology and they are currently available only in silicon. In the silicon drift detector (SDD) the biasing rings and the back contact shape the electric field inside the device, as can be seen in radial structure in Fig.1(e). The electric field drifts the electrons released in an interaction (anywhere in the active volume) into a valley away from the recombining interfaces, and then drives them toward a single electrode. The holes are collected by the biasing rings. Most of the signal current at the collecting electrode would be from conduction current, rather than induced current as in the planar depleted devices. The detector is a unipolar device since only the electrons contribute to the output signal. This device needs fine biasing tuning, but yields superb energy resolutions and counting rates even around 300 K. Since there is no depletion region around the anode—the capacitance of SDD is very low. Furthermore, in a planar pad detector if several interactions occur within a short time interval, they all contribute to the induced current significantly at the same time, whereas in SDD such events can be easily resolved unless they also occur at the same radial distance from the anode. The electrons generated by different events drift in distinct bunches and contribute conduction current only upon arrival. This results in a higher counting throughput (in terms of rate). In the more modern version of the SDD，the first stage transistor is integrated into the device by the anode.

The float gate structure shown in Fig.1(f) was only recently presented and it is based on a common float-gate memory technology. The float-gate is pre-charged, while each even reduces the amount of charge either by opposite charging or by assisting emission.

2.4 Group IV detectors

Some detector related properties of group IV detector semiconductors are summarized in Tab.1.

Tab.1 Partial summery of electrical properties of group IV semiconductors for detector application

Silicon is the most studies material in the world of electronics. It has many favorable electronical and mechanical properties. It’s bandgap of 1.12 eV enables room temperature operation for most applications with sufficiently low carrier generation rate, which yield reasonable leakage currents at junctions. The material facilitates fairly easy and high level dopings of both, n and p types. The material is usually grown by Czochralski[2-4]or float zone (FZ)[5]techniques, and can be grown in a wide range of specific resistivities form mΩ×cm to dozens of kΩ×cm. The availability of high quality local oxide tops the list of advantages. Silicon has indirect bandgap which makes it unfavorable for optical applications; however there were certain attempts to evaluate it[6-8]. Part of the UV spectrum is considered within the “ionizing radiation” range, and recently high efficiency UV detectors were reported on silicon[9]. Since the UV absorption range in silicon is in the nanometer range, the authors introduced high electric field at the incident surface using ultrathin p+-i-n junction which yielded avalanche gain of 2 800.

Silicon detectors are widely used for high energy physics experiments, where they are responsible for particle tracking[10-13]. When photons penetrate the detector they leave no trail until the point of interaction (if it takes place inside the detector volume). Other particles (with non-zero rest mass) start colliding the media from the moment of entry, leaving a trail of generated (free) charge carriers as well as ionized-, excited-, and displaced- atoms. Thus, for MIPs (that deposit only a small fraction of their energy in the detector) the signal strength is proportional to the active layer thickness. Typically 200～400 μm thick devices are needed for reliable operation. Radiation hardness of such detectors has been intensively studied for years[14-15].

Since the atomic number of silicon is 14, its absorption of high energyphotons(x- and gamma- rays) is very limited. However, for some applications such as low energy photon energy spectroscopy, ion detection and spectroscopy, low energy proton spectroscopy, silicon can be used. Depending on the application silicon is commonly used in Schottky barrier[16-18], asymmetric p-n junction[19-23], and p-i-n configurations[24-26]. Since silicon technology is highly developed, more complex structure can be produced, such as silicon drift detector (SDD)[27], and floating gate detector (and dosimeter)[28].

Full depletion of a few hundreds of microns is easily achievable using high resistivity FZ grown wafers[29-31]. However, higher photon energies require several mm thickness,thus yet lower carrier concentrations are needed. There are two common methods to reduce the effective carrier concentration in silicon: neutron transmutation doping (NTD)[32-34], and lithium drift[35-38]. In both methods medium resistivity (few kΩ×cm) FZ grown p-type silicon is used as a starting material. Typically silicon consists of three isotopes:28Si-92.2%,29Si-4.7%, and30Si-3.1%. In NTD method the ingot of low doped p-type silicon is irradiated by thermal neutrons, which react with the stable30Si isotopes turning them into unstable31Si, which decay with 2.62 h half-time into (β-)31P (emitting antineutrino). The gradual and nearly uniform (in terms of volume distribution) conversion of silicon isotopes into phosphorus is carefully monitored. Thus, at some point the initially p-type material becomes compensated by addition of phosphorus, opposite type shallow level. The rate of conversion can be controlled by controlling the thermal neutron flux. Around the point of type inversion the material is closest to intrinsic in terms of free carrier concentration.Lithium atoms form interstitial donors in silicon and germanium,when they are massively diffused into p-type silicon an n+-p junction is formed. The method was first introduced in 1960 by Pell[36-38]for diffusion studies, and 2 years afterwards adopted by Freck and Wakefield for Si(Li) detectors[39]. After initial lithium diffusion the n+-p junction is reverse biased and the device is heated to increase lithium ion mobility. As a result the lithium ions start drifting into the p region. As the p region is being gradually compensated，the effective junction position is moving and with it the maximum electric field region. In addition，the lithium ions can bond to boron atoms thus neutralizing the p-type dopants. As the drift reaches the back contact,the effective junction disappears dramatically increasing the current, which is used as a signal to end the drift. The compensating lithium ions are mobile even at room temperature, thus Si(Li) detectors are only used with cooling in order to prevent out-diffusion/drift of the lithium. Needless to say that only cold processes are applicable to this material after the compensation (e.g., no doping and junction formation).In 1983，a more complex and advanced detector structure was proposed by Gatti and Rehak[27]—the silicon drift detector (SDD). The concept, schematically shown in Fig.1(e), gained considerable popularity in the following years[40-42].Another interesting configuration similar to floating gate RAM cell was recently introduced by Pikhay[28]. The principle configuration is shown in Fig.1(f). Initially the floating gate (FG) is charged from the control gate (tunneling by applying high electric field). The holes that are released in the active volume due to particle interaction drift toward the floating gate and neutralize the excess electrons there, it is also assumed that the hot carrier travelling through the floating gate would excite the pre-charged electrons there and facilitate their emission. Both mechanisms should reduce the negative FG charge, which affects theId-Vgcurves. The device is resettable by gate tunneling pulse, and it measures the particle fluence; however, the flux can be deduced from dI/dtderivative.