Nuclear Inst.

and Methods in Physics Research, A 960 (2020) 163663

Contents lists available at ScienceDirect

Nuclear Inst. and Methods in Physics Research, A

journal homepage: www.elsevier.com/locate/nima

A first principle method to simulate the spectral response of CdZnTe-based

X- and gamma-ray detectors
Manuele Bettelli a ,∗, Nicola Sarzi Amadè a , Davide Calestani a , Bruno Garavelli b , Pietro Pozzi b ,
Daniele Macera b , Luca Zanotti b , Carlo Andrea Gonano b , Matthew C. Veale c , Andrea Zappettini a
a
IMEM/CNR, Parco Area delle Scienze 37/A, Parma 43100, Italy
b
Xnext s.r.l., Via Adelaide Bono Cairoli 30, Milano 20127, Italy
c Rutherford Appleton Laboratory, Chilton, Oxfordshire OX11 0QX, UK

ARTICLE INFO ABSTRACT

Keywords: Detectors based on compound semiconductor materials like CdTe and CdZnTe are more susceptible to defect-
CdZnTe related spectral distortions than elemental semiconductors like Si or Ge. During the design process of new
Detector simulation detectors based on these materials it is crucial to consider the effect of these distortions on the detector
Monte Carlo simulation
performance. Due to the diverse range of application areas in which these detectors may be used, the detector
Signal induction
geometry must be selected to match the desired application of the device. For those requiring the detection
of photons across a broad energy range (1 – 1000 keV), the detector design must account for a variety of
different interaction processes.
The simulation framework presented in this paper includes all the physical processes involved in the
formation of the detector signal, from the radiation absorption mechanisms to the influence of the electrode
geometry. A simulation system based on first principle calculations is used which consists of a Monte Carlo
simulator, a Finite Elements Method (FEM) calculator and numerical computation software. The framework
simulates the radiation–semiconductor interaction, the charge carrier transport and the role of the electric field
and weighting field in signal induction on the electrodes. This tool allows to simulate the entire experimental
arrangement including the use of attenuators, collimators and scattering surfaces. The ability to accurately
simulate the detector response to radiation and its surroundings provides a powerful tool for the realization
of a new generation of detector systems.
In order to validate the simulation framework, CdZnTe-based detectors with several contact geometries
have been modelled and the output of the simulations have been compared to experimental data. A comparison
between the simulated and measured responses demonstrate the power of this technique.

1. Introduction One of the key advantages of compound semiconductor materi-

als like CdTe and CdZnTe over traditional elemental semiconductors
Ionizing radiation detectors are today employed in a diverse range like Si and Ge is their wide bandgap and resulting high resistivity
of applications: these include environmental surveillance (background (>1 × 109 Ω cm) which allows them to be easily operated at room
radiation monitoring), homeland security, astrophysics (gamma-ray temperature. Moreover, its density provides high efficiency across a
space-investigation), non-destructive testing (in-line checks for food or
broad range of photon energies 1–1000 keV. On the other hand, the
products quality), and medical imaging.
greater crystallographic complexity of the binary and ternary semi-
Over the last few decades the prominence of semiconductor-based
conductors means they are more susceptible to defect-induced spectral
devices within the X-ray and 𝛾-ray imaging communities has grown
replacing traditional scintillator-based systems. Unlike scintillators, distortions compared to Si or Ge. Understanding the properties of these
semiconductor-based detectors directly convert the ionizing radiation compounds is crucial during the designing process of new detectors.
into an electrical signal leading to devices with much higher energy The engineering of the crystal dimensions and the contact geometry
resolution. In addition, the ability to segment detector electrodes allows can be used to optimize the detector final performance in order to limit
to produce devices with much higher spatial resolution [1–3]. the effect of charge trapping and charge sharing effects. In particular,

∗ Correspondence to: Parco Area delle Scienze 37/A, Parma 43100, Italy.
E-mail address: [email protected] (M. Bettelli).

https://doi.org/10.1016/j.nima.2020.163663
Received 20 December 2019; Received in revised form 14 February 2020; Accepted 15 February 2020
Available online 18 February 2020
0168-9002/© 2020 Elsevier B.V. All rights reserved.
M. Bettelli, N.S. Amadè, D. Calestani et al. Nuclear Inst. and Methods in Physics Research, A 960 (2020) 163663

the electrode fabrication technology and design play a key-role: if not values are both extrapolated on a uniform 3D square mesh with about
optimized, this may result in poor performance even when a high- 5 μm steps, depending on the detector geometry and dimensions. The
quality material is used. The optimal electrode geometry will depend weighting field depends only on contacts geometry and is defined by
on the final application: energy range, photon flux and collecting angle the Shockley–Ramo theorem [14] and it describes the charge induced
are important factor which must be considered. on a given electrode as a charge carrier drifts through the device. The
In this paper a simulation framework based on first principle calcu- weighting potential on the collecting electrode is calculated by setting
lations is presented. The tool can simulate the main physical processes its potential to 1 V whereas the potentials of all other electrodes are
involved in the detection of X- or 𝛾-rays, from the radiation absorption set to 0 V. The outputs of this block include: (i) the X, Y and Z lattice
mechanisms to the influence of the electrode geometry on the signal coordinates; (ii) the electric potential evaluated at each grid node
generation process. The capability to predict the detector response and and (iii) the weighting potential at each grid node. The electric and
to engineer it before beginning the costly device fabrication and testing weighting fields are successively calculated as the gradient of respective
process has the potential to drastically improve the development of the potentials.
next generation of detector systems [4–8]. Thanks to the flexibility of the software, it is possible to simulate
The framework is composed by three main blocks: the radiation– 3D detectors with any type of electrode geometry and configuration. In
semiconductor interaction is simulated by Monte Carlo methods in each simulation, the crystal dimensions, the location of the electrodes
GEANT4; the electric and weighting fields are simulated by finite and their respective electric potential can be defined. This package
elements method with COMSOL Multiphysics; finally, the charge carrier allows to insert a spatial charge within the crystal which, in case of
transport and signal induction are calculated in MATLAB environment. planar detector, leads to a linearly increasing/decreasing electric fields
A comparison between simulated and measured spectra have been along the detector thickness. We assume a stationary conditions in
carried out in order to validate the power of this tool. CdZnTe-based which the electric field is not distorted by external factors. In fact,
detectors with different dimensions and contact geometries have been CZT is not affected by bias induced polarization effects which would
simulated and the agreement with the respective experimental mea- otherwise create a dead layer region near one of the contacts due to
surements has been quantified. an accumulation of positive or negative charge. Moreover, high-flux
grade CdZnTe can hold high photon fluxes which translates into the
2. Simulation methods and results absence of radiation induced polarization [13]. This assumption has
been proven experimentally with different techniques [15–18].
Several simulation techniques reported in the literature are based on The choice of detector geometry strongly depends on the properties
semi-empirical approaches [9–12]. Unfortunately, even if good results of the semiconductor used as well as on the requirements of the final
can be achieved using these techniques, they require specific device application. For example, in the field of X-ray and 𝛾-ray spectroscopy
characterization measurements to calibrate the parameters of the cho- researchers have developed different detector geometries which reduce
sen model (such as electric and weighting field approximation or the effects of severe hole trapping in CdTe and CdZnTe materials (single
parametrization of tailing effect due to hole trapping). By comparison, polarity charge sensing) [14,19].
the proposed simulator does not require preliminary calibration and In the following sections the most typically used detector geometries
the spectroscopic response of the detector can be calculated without are described. A qualitatively and almost size-independent behaviour of
producing prototype devices and collecting experimental data to train electric and weighting potentials for each geometry is shown in Fig. 2
the model. Simulator requires parameters that are related to constituent to provide a general overview of dominant trend.
material (transport properties and spatial charge), these parameters are
typically provided by material sellers. A simulation-led design approach 2.1.1. Planar–planar electrodes
that allows a variety of electrode geometries and dimensions to be In the planar–planar configuration, two large continuous electrodes
tested without the need for device fabrication and testing has the cover the two biggest faces of the semiconductor crystal as illustrated
potential to deliver large benefits in terms of reduction in R&D costs. in the first row of Fig. 2.
Our approach is based on the transport of charges generated by Readout electronics is connected to one of the electrodes (typically
the interactions between the X-ray and 𝛾-ray photons with the semi- the anode), while the bias voltage is provided to the other electrode
conductor material. The movement of these charges in the detector (cathode). In such a geometry, the weighting potential is a linear
material induces a signal on the collection electrodes as described function of depth within the detector and has a value ranging from 0
by the Shockley–Ramo theorem [13]. The charges collected at each to 1. The resulting weighting field is uniform over the entire detector
electrode are calculated by integrating the simulated signals over time thickness (L). In this simple geometry, the weighting and electrical
and adding random electronic noise with a Gaussian distribution as a fields shapes are identical and the value of the electric field is equal
first approximation; additional noise sources, such as the detector elec- to V/L, where V is the applied bias.
tronics, can also be introduced in this step of the simulation chain. This
process is then repeated for several photon interactions and the charge 2.1.2. Virtual Frisch-grid electrodes
detected for each pixel is recorded in an energy spectra per pixel. The If trapping can be ignored during the drift time, as is usually the
simulator workflow is shown in Fig. 1; each block is explained in detail case in a high purity germanium detector (HPGe), the induced charge
below. on the anode does not depend on the depth of interaction. In the case of
CdTe and CdZnTe, the poor transport features related to the holes cause
2.1. Simulation of electric and weighting fields a strong photon interaction-depth dependency in the overall detector
response. Interactions occurring close to the anode will result in signals
The electric and weighting fields are calculated by using the ‘‘elec- that are dominated by the holes contribution, thus leading to a signifi-
trostatics’’ package of COMSOL MultiPhysics. Depending on the ge- cant charge losses whereas interactions occurring close to the cathode
ometry and the required accuracy, this tool automatically creates an are dominated by electrons and are less affected by charge loss. One of
adaptive mesh which is denser in the high-gradient region and coarser the first single polarity charge sensing techniques implemented in gas
elsewhere. Usually, the former regions are located near the anode in detectors was developed by Frisch [20]. In this geometry an additional
the case of small pixels/strips. Adaptive mesh builder is helped setting grid contact between the cathode and anode has been introduced to
minimum (50 nm) and maximum (100 μm) elements size, number of reduce the significance of the charge induced by drifting ions while
volume elements used by the simulator is automatically calculated they are far from the anode. Since the fabrication of such a grid contact
and typically varies from 1 to 5 106 . Electric and weighting potential is not feasible in semiconductor detectors, a virtual grid is usually

2
M. Bettelli, N.S. Amadè, D. Calestani et al. Nuclear Inst. and Methods in Physics Research, A 960 (2020) 163663

Fig. 1. Simulator concept scheme.

realized: the lateral surfaces of the detector are firstly covered with an potential is very similar to that of the virtual Frisch-grid detector where
insulating layer and then they are enveloped in a metallic sheet that carriers induce a significant charge in the detector electronics only
plays the role of the Frisch grid. In this configuration, called virtual when they drift close to electrodes ‘‘A’’ and ‘‘B’’.
Frisch-grid [21], the lateral metal layer is close to the anode in order
to minimize the region where carriers induce a signal. The resulting 2.1.4. Drift-strip electrodes
weighting potential is almost zero from the cathode to the end of the
Semiconductor detectors using strip electrodes were developed for
metallic layer and strongly increases near the anode (second row in
two-dimensional position sensing [25,26]. This geometry is also suit-
Fig. 2). The charge induced by the drifting carriers on the readout
able for single-polarity charge sensing by reading out signals from
electrode is thus close to zero until they enter the region close to the
individual strip electrodes [24–26]. The weighting potential of a spe-
anode.
cific strip is calculated applying 1 V on a given strip and then setting all
Bolotnikov et al. [22] observed that virtual Frisch-grid detectors
the other electrodes (strips and the cathode) to 0 V. If the strip-pitch is
with rougher lateral surfaces yielded better spectroscopic performance,
much lower than detector thickness, the resulting weighting potential
even though they had higher electronic noise and leakage current. It
is close to zero in most of the detector volume, except very close to the
was proposed that the higher surface leakage current resulted in a more
uniform electric field distribution inside the detector, thereby improv- strip (fourth row in Fig. 2).
ing the charge collection efficiency. Similar effects were also obtained Instead of using a strip detector for performing a two-dimensional
when the crystal edges were treated with a chemical process [22]. In position sensing, a small modification in its electrodes bias allows
order to replicate the response of realistic detectors, the simulations it them to be used as a drift detectors for high resolution spec-
presented here include higher leakage currents on the lateral crystal troscopy [27,28]. The electric potential shown in the fourth row of
surfaces. This effect is obtained by imposing a constant voltage gradient Fig. 2 is obtained by using a stepped bias voltage profile, in which
on detector edges. both the cathode and the outermost strips are held at the maximum bias
voltage while the other strips are set to bias voltage which progressively
2.1.3. Coplanar grid electrodes reduced until 0 V is reached in correspondence to the central collecting
A different method of obtaining single-polarity charge sensing in strip. The resulting electric potential drives charge carriers towards the
semiconductor detectors was implemented by Luke [14,23] who em- central collecting electrode regardless of their initial point of creation.
ployed the use of coplanar grid electrodes. The concept of a ‘‘coplanar
grid’’ is illustrated in the third row of Fig. 2. In this electrode design, 2.1.5. Quasi-hemispheric electrodes
the anode consists of two sets of interdigitated grids each with its Hemispherical detectors are obtained by using a small collecting
own readout electronics which are referred to as grids ‘‘A’’ and ‘‘B’’ anode at the focus point of the hemispheric cathode. The weighting
in the discussion that follows. A voltage difference between grid ‘‘A’’ potential of this small anode is calculated by applying a unit potential
and ‘‘B’’ is applied so that the selected carriers are always collected by
on the anode and grounding the cathode. The resulting weighting
one of the electrodes. If ‘‘A’’ is the collecting electrode, its weighting
potential is almost null within most of the detector volume but rapidly
potential (Vw A ) is calculated by setting its to 1 V, while the potentials
rises to 1 as the carriers approach the anode. Analogously to the ‘‘drift-
of the cathode and grid ‘‘B’’ are both set to 0 V [24]. The weighting
strip’’ configuration, the electric field within the detector results in
potential map in the third row of Fig. 2 has been obtained by solving the
the drift of charge carriers towards the small anode regardless of the
Laplace’s equation. Similarly, the weighting potential of electrode ‘‘B’’
position of their original creation. The use of this geometry allows the
(Vw B ) can be calculated by repeating the process but this time setting
creation of detectors with large active volumes for high-energy X-ray
the potential of electrode ‘‘B’’ to 1 V, with the potentials of the cathode
and 𝛾-ray spectroscopy while limiting at the same time the degraded
and grid ‘‘A’’ both set to 0 V.
contribution related to the holes low-mobility.
Fig. 3 shows the behaviour of the weighting potential along a
given cross-section through the detector (see the dotted line in the third With CdTe and CdZnTe the fabrication of hemispheric detector
row of Fig. 2) which is perpendicular to the electrode surfaces and crystals is so technologically challenging that such a geometry is pre-
intersectings with the collecting electrode ‘‘A’’. The cross-section of the vented, in that, a proper instrumentation would be required. The
weighting potential as observed in the neighbouring strip of electrode ‘‘quasi-hemispheric’’ detector geometry was thus proposed in order to
‘‘B’’, is also shown in Fig. 3. It can be noted that these two cross- overcome this limitation [29]. The idea is to realize the hemispheric
sections, Vw A and Vw B , are almost identical for 0 < z < L-P (where z detector on a parallelepiped-shaped CZT block, as shown in the fifth
is the linear coordinate along the line, P is the period of coplanar grid row of Fig. 2, by maintaining the same anode shape and by extend-
electrodes and L is the sample thickness), whereas for higher z values ing a continuous metallization on the lateral surfaces. The physical
Vw A approaches 1 and Vw B rapidly drops to 0. manufacturing of such a detector is technologically feasible.
In this instance, single-polarity charge sensing is implemented by The electric and weighting potential shapes obtained in this way are
reading out the difference between the signals induced on the collecting very similar to that of the hemispheric cathode configuration. Electrical
electrode ‘‘A’’ and the non-collecting electrode ‘‘B’’ (Vw A - Vw B ) as current flowing between the pixel and the lateral metallized surfaces
shown by the black line in Fig. 3. The resulting differential weighting ensures a more uniform electric field distribution.

3
M. Bettelli, N.S. Amadè, D. Calestani et al. Nuclear Inst. and Methods in Physics Research, A 960 (2020) 163663

Fig. 2. Qualitative results of the simulation on different kind of example detectors: contacts type and detectors shape are represented in the first column; weighting potential maps
of a cross-section of each detector are in the second column; normalized electric potential maps of a cross-section of each detector are in the third column (further details in the
text). Dotted line for coplanar configuration is related to Fig. 3 plot.

2.1.6. Pixel electrodes degrading the overall system performances in terms of a very high
Pixelated detectors are exploited in position-sensitive imaging appli- readout noise. The presence of a guard ring, which must be biased at
cations [30,31]. Single-polarity charge sensing is achievable when the either the same or very-close-to voltage as that of the pixels, instead,
pixel geometric dimensions are much smaller than the device thickness makes the surface leakage current to be collected by the guard itself.
(the so-called small-pixel effect) [32]. With this geometry the charge This prevents such current to be collected by the pixels front-end
induced on a given pixel is small when the charge cloud is far from the electronics, a beneficial situation which allows to significantly reduce
pixel but rapidly increases as it moves close to the electrode (sixth row the readout noise and thus to improve the energy resolution of the
of Fig. 2). detector, especially for pixels at the edge of the array.
Typically, pixel detectors also include a guard-ring electrode sur-
rounding the whole array. The role of such additional electrode is to 2.2. Simulation of photon sources and radiation–matter interaction
minimize the contribution, to the overall pixel response, introduced by
the detector edges. In fact, lateral detector surfaces present a much In the second simulation step the photon source and the radiation–
lower resistivity than the bulk material, so that, due to the intense matter interactions inside the crystal volume are simulated with
electric fields applied to the device for making it working properly, a GEANT4 [33]. This toolkit makes use of large database containing
very high surface leakage current typically flows through them. This physical models and experimentally measured cross-sections to re-
current is collected by the detector readout electrode, thus strongly produce the absorption mechanisms of particles within matter via

4
M. Bettelli, N.S. Amadè, D. Calestani et al. Nuclear Inst. and Methods in Physics Research, A 960 (2020) 163663

Fig. 3. Weighting potential calculated along the dotted line showed in Fig. 2 (coplanar Fig. 4. Representation of carrier transport and signal induction in a randomly chosen
grid, weighting potential map) for electrode A and B. The black line represents the moment of carrier travel simulation. The actual position of carriers and their route
difference between the two. (For interpretation of the references to colour in this figure are respectively represented with dots and routes. The colour and intensity of the
legend, the reader is referred to the web version of this article.) line represents the electric current induced on the anode. Simulation concerns only
the electron transport in a quasi-hemispherical detector. (For interpretation of the
references to colour in this figure legend, the reader is referred to the web version
of this article.)
Monte Carlo simulations. We used the libraries G4EmStandardPhysics
and G4EmLivermorePhysics which are specialized in electromagnetic
interactions between photon and matter in the energy range from few calculates the induced current for both electrons and holes via the
keV up to tens of MeV. These packages allow to consider the following Ramo–Shockley’s theorem. Carriers positions, trail paths and induced
physical processes: signals are all represented in Fig. 4. In this plot, the actual position of
• For photons: gamma-photon conversion to electron/positron pair, the charge clouds and their routes are respectively shown (the colour
Compton scattering, Rayleigh scattering, photoelectric effect; indicates the intensity of the induced current).
By integrating the induced current on a given electrode as a function
• For both electrons and positrons: ionization, Bremsstrahlung,
of time it is possible to determine the net induced charge. Theoretically,
Multiple scattering, annihilation (positrons only).
this is the correct value to use to get the spectrum but, in order to
The tracking of secondary electrons or photons is possible, allowing compare the simulated spectrum with the experimentally measured
to estimate the initial physical dimension of the charge cloud (see the spectrum, it is also necessary to add the contribution of the readout
next subsection). The distance cut-off limit for secondary particles in electronics to the final detector signal (pulse shaping). The developed
‘‘multiple scattering’’ process type has been set to 500 μm. A planar software accounts for this contribution by adding its specific behaviour
‘photon gun’ is defined whose size is identical to that of the detector instead of a simple current integration.
sensor (except for synchrotron beam simulations, where the actual For example, by considering only the electron contribution, the
beam dimension has been used) and it is located at a distance of 50 cm. induced charge due to a charge initially located at 𝑟𝑖0 is calculated as
Photons trajectories are perpendicular to the plane of the detector and follows. The trajectory is obtained by using:
their spatial distribution on the face of the detector is assumed to be 𝑡 𝑡 ( )
𝑟𝑖 (𝑡) = 𝑟𝑖0 + 𝑣(𝑡′ )𝑑𝑡′ = 𝑟𝑖0 + 𝜇𝑒 𝐸 𝑡′ 𝑑𝑡′ (1)
uniform. Eventual collimators or additional passive layers attenuating ∫0 ∫0
the radiation can be implemented in this step as well. The geometric di-
where 𝑣(𝑡′ ) is the carrier velocity, 𝜇𝑒 is the electron mobility and 𝐸(𝑡′ )
mensions of the detector and constituent materials are defined as in the
is the electric field experienced by the carriers in the time 𝑡′ . We
previously described step. Both the released energy and coordinates of
assume that 𝜇𝑒 is independent from the electric field since its saturation
each either photon or electron interaction within the detector sensitive
occurs at value greater than 15 000 V/cm [34]; typical field strengths in
volume are saved into a CSV file.
CdZnTe detectors are of the order of 5000 Vcm−1 . We have assumed the
Different radiation sources such as radioactive isotopes (241 Am,
57 Co, 137 Cs), X-ray tubes or synchrotron light have been simulated. The electric field is not time-dependent per se, but charges that travel inside
the detector experience an electric field value that actually changes
number of events generated in a single simulation was typically in the
accordingly with that specific point in which, at any time instant, they
order of 106 .
are currently passing. For this reason, the electric field 𝐸(𝑡′ ) felt by
carriers is time dependent. The induced current on a given electrode
2.3. Simulation of charge carriers transport and signal generation
at time t is calculated using the Ramo–Shockley’s theorem:
( ) ( )
The carriers drift is simulated in the third step of the simulator 𝑡 ( ) ( ) 𝑡 ( )
𝐼 𝑖 (𝑡) = 𝑄0 𝑒𝑥𝑝 − 𝜇𝑒 𝐸 𝑟𝑖 (𝑡) ⋅ 𝑊 𝑟𝑖 (𝑡) = 𝑄0 𝑒𝑥𝑝 − 𝑣(𝑡) ⋅ 𝑊 𝑟𝑖 (𝑡)
(labelled ‘‘3a’’ in Fig. 1) and the resulting readout current signal is 𝜏𝑒 𝜏𝑒
so produced. The simulator assumes a charge carriers cloud starting
(2)
point, within the detector volume, which is coincident to the generated ( 𝑖 )
charges interaction position calculated in the previous step (‘‘2’’). The where 𝜏𝑒 is the electron lifetime, 𝑄0 is the initial charge and 𝑊 𝑟 (𝑡)
charge carriers are then successively drifted according to the electric is the weighting field. The 𝜏𝑒 value used is obtained experimentally
field obtained in step (‘‘1’’). The software calculates trajectories of (see Table 2). Thus, it is an average value which considers both the
both carriers types by means of an ODE (Ordinary Differential Equa- effect of crystal defects and secondary phase like Te inclusions, a well-
tion) function. Once motion equations are generated, the software known recombination centre. Assuming a gaussian distribution of the

5
M. Bettelli, N.S. Amadè, D. Calestani et al. Nuclear Inst. and Methods in Physics Research, A 960 (2020) 163663

charge density, the broadening of the carrier cloud as a function of Table 1

Summarization of tested detectors and photon source.
the time due to the charge diffusion and the Coulomb repulsion can
be calculated with the following equation: Geometry Dimension Photon source
Large pixel detector Detector 4 × 4 × 2.8 mm3 241 Am
𝑡
𝑑𝑡′ 𝜇𝑘𝐵 𝑇 Pixel 2 × 2 mm2 nuclear source
𝜎 2 (𝑡) = 𝜎02 + 2𝐷𝑡 + 2𝛼 √ with 𝐷 = and
∫0 𝜎 2 (𝑡′ ) 𝑞 Small pixel detector Detector 5.5 × 19.2 × 2.5 mm3 25 and 50 keV
( )√ Pixels 0.2 × 0.2 mm2 monochromatic
3𝜇𝑁𝑞 5 synchrotron radiation
𝛼= (3)
4𝜋𝜀 15
Drift-strip detector Detector 20 × 4.5 × 6 mm3 137 Cs

where 𝜎 2 (𝑡) is the variance at time 𝑡, 𝐷 is the diffusion coefficient, 𝜇 Strips 0.25 mm nuclear source
Gap 0.55 mm
is the mobility, 𝑘𝐵 is the Boltzmann’s constant, 𝑇 is the temperature,
𝑁 is the number of charges, 𝑞 is the absolute electron charge and 𝜀
is the electric permittivity of CdZnTe. The initial cloud dimension 𝜎 2 0
has been obtained via GEANT4 simulations in which secondary inter- to validate the simulator. In the following paragraphs, the characteris-
actions of a photoelectron of a given energy are tracked (an example is tics of the tested devices, together with their relative, readout systems,
shown in Fig. 5). The variance has been obtained by fitting the energy are described. Table 1 summarizes the different geometries which have
( )
weighted interaction positions with a gaussian model. The 𝑊 𝑟𝑖 (𝑡) been tested.
value used to calculate the induced signal is obtained by weighted- The comparative criteria are different for each spectrum, since the
averaging the weighting field on the whole charge cloud volume. This distortions which affect the energy spectrum and the possible sources
approach is powerful since it allows to include the effect of the cloud of mismatch depend on several parameters such as the energy range,
size on the current transient which becomes important at high energy the physical surroundings and the electronic noise. These criteria are
as shown in Fig. 5. However, in most cases the geometric dimensions listed below (all spectra are normalized by area):
of the electrodes are much bigger than the charge cloud and typical
times of flight in CdZnTe are about few tens of ns at most. Then, • Normalized Standard Deviation (NSD)
√
under a certain threshold the cloud can be considered as punctiform ( )2
∑ 𝐸𝑖 − 𝑆𝑖
(see Fig. 5). In the case of pixelated detectors this value is around 𝜎=
200 keV whereas for drift-strips/hemispheric detectors the threshold is 𝑁
around 1 MeV. In fact, the electric field lines convergence counteracts where 𝐸𝑖 and 𝑆𝑖 are the number of counts of the experimental
the cloud spreading allowing a better collection efficiency. However, and simulated spectra in the 𝑖th bin, respectively, and 𝑁 is the
cloud splitting cannot be considered since only the trajectory of the total number of bins (8192).
cloud barycentre is calculated. • Area under the Compton edge (AC). Used for Caesium spectra
Finally, the induced charge is obtained by integrating Eq. (2) with ∑
𝐴𝐶𝐸 = 𝐸𝑖
respect to time: 𝑃 ℎ𝐸<500 keV
( ) ∑
𝑇 𝑂𝐹
𝑡 ( ) 𝐴𝐶𝑆 = 𝑆𝑖
𝑄𝑖𝑛𝑑 = 𝑄0 𝑒𝑥𝑝 − 𝑣(𝑡) ⋅ 𝑊 𝑟𝑖 (𝑡) 𝑑𝑡 (4)
∫0 𝜏𝑒 𝑃 ℎ𝐸<500 keV

where 𝑇 𝑂𝐹 is the carrier time of flight that is the time required for the where 𝑃 ℎ𝐸 is the photon energy
charges to complete their drift from the interaction position to the col- • Photopeak/Compton edge ratio (P/C)
lecting electrode. Any contribution from the electronics components, if • Area under the fluorescence and escape peaks (AFE). Used for Am
known, can be added in this simulation step. More accurate simulation spectra
∑
results can be indeed achieved in case of knowledge about the transfer 𝐴𝐹 𝐸𝐸 = 𝐸𝑖
function that converts the induced charge signals into the read energy. 𝑃 ℎ𝐸<40 keV
∑
This block has been implemented in MATLAB. 𝐴𝐹 𝐸𝑆 = 𝑆𝑖
𝑃 ℎ𝐸<40 keV
2.4. Spectrum reconstruction
3.1. Large-pixel detector

As a final step (labelled ‘‘3b’’ in Fig. 1) the energy spectra for each The simulations were initially validated considering the case of
electrode is built starting from the induced charges calculated before. a large single-pixel detector. The sample was realized with Redlen
Events generated by the same photon are added together because they standard spectroscopic grade CdZnTe material. Gold contacts were
occur on a timescale of few picoseconds from each other and can be fabricated on the CdZnTe crystal using an electroless deposition pro-
treated as a single event. For example, in the case of the generation of cess [35,36]. The dimension of the cathode electrode is 4 × 4 mm2
Cd and Te fluorescence photons, the charge induced by the interaction and the anode consists of a central pixel of 2 × 2 mm2 surrounded by
of these photons is added to the principal photon induced charge. In a a guard ring. The CdZnTe crystal thickness is 2.8 mm and the pixel-
pixelated detector, charge induced by fluorescence photons interacting guard ring gap is 50 μm. The spectroscopic performance of detectors
far from the volume below the pixel is close to zero and then the with these same dimensions have already been previously characterized
contribution of these photons is lost, thus producing the escape peaks. by Abbene et al. [37,38]. The detector was biased at −800 V and the
Finally, the electronics noise is added to the simulated spectrum as cathode has been irradiated with a 241 Am 𝛾-ray source (∼60 keV).
a gaussian broadening whose extent depends on the specially adopted Detector temperature was 25 ◦ C. The detector electronics used for
electronic read-out chain. This value has been evaluated for all the these measurements consisted of a charge sensitive preamplifier and
investigated test-cases for comparison with the experimental data col- a digitizer that captured the pulse shapes measured at the anode pixel.
lected with the real detectors (Fig. 6). Digital pulse shape processing was used to shape the signals from the
charge sensitive preamplifier. Electronic noise was 2.1 keV.
3. Validation of the model with experimental results The comparison between the experimental and the simulated en-
ergy spectrum is shown in Fig. 7. The simulation was carried out for
Detectors with extremely different geometries were simulated and spectroscopic grade CdZnTe material using the values for the charge
the results compared with those obtained from measurements in order carrier transport properties listed in Table 2 [39]. Both the simulated

6
M. Bettelli, N.S. Amadè, D. Calestani et al. Nuclear Inst. and Methods in Physics Research, A 960 (2020) 163663

Fig. 5. Left: example of interaction positions of a photoelectron and its secondary particles. Right: time evolution of carrier cloud size generated by photoelectrons in the range
50–1000 keV due to Coulomb repulsion and charge diffusion.

Fig. 6. Left: spectrum obtained by adding together the deposited energy of multiple interactions of each single photon (only Geant4 simulation). Centre: spectrum after the
electrode charge collection. Right: spectrum after the gaussian broadening.

and the measured spectra demonstrate a high degree of agreement and,

consequently, a small NSD value of 0.003 calculated across the energy
range 0–80 keV. The experimental and simulated AFE values are 0.596
and 0.598, respectively, which corresponds to a difference of 0.3%.

3.2. Small-pixel detector

The second validation of the simulation framework was carried

out for a 2D pixelated detector. This detector was fabricated from
2.5 mm thick standard Redlen spectroscopic grade CdZnTe material.
The cathode area was 5.5 × 19.2 mm2 and the anode consisted of an
array of 20 × 75 pixels having a pitch of 250 μm with a pad side width
of 200 μm and an inter-pixel gap of 50 μm. The detector was flip-chip-
bonded directed to the spectroscopic X-ray imaging Hexitec ASIC [40].
The measurement was carried out at 28 ◦ C, further details about the
electronics setup are reported by Jones at al. [40]. The detector testing
was performed at the Diamond Synchrotron (UK) using a collimated
beam of monochromatic X-rays. The 10 × 10 μm2 beam was collimated
using two sets of tungsten slits and was aligned with the pixel centre. Fig. 7. Comparison between experimental and simulated energy spectra of 241
Am
Two different monochromatic X-rays of 25 and 50 keV (below and nuclear source.
above the K-edge of Cd and Te) were used. The electronic noise is 0.7
keV which corresponds to a FWHM of 2.7% and 1.5% at 25 and 50
keV, respectively. The comparison between the spectra measured using

7
M. Bettelli, N.S. Amadè, D. Calestani et al. Nuclear Inst. and Methods in Physics Research, A 960 (2020) 163663

Table 2
Overall table containing simulation parameters for all experiments (Carriers parameters measured by Thomas et al. for standard Redlen CdZnTe
[36]).
Parameter Large-pixel detector Small-pixel detector Stripes detector
Electron mobility 1000 1000 cm2 /Vs 1000 cm2 /Vs
Electron lifetime 11 μs 11 μs 11 μs
Hole mobility 88 88 cm2 /Vs 88 cm2 /Vs
Hole lifetime 0.2 0.2 μs 0.2 μs
Cathode bias voltage −800 V −750 V −450 V
Drift-strips bias voltages Not present Not present −450 V ; −300 V ; −150 V
Anode Bias Voltage 0 V 0 V 0 V
241 137
Photon source Am 25 and 50 keV monoenergetic Cs
beam, 10 × 10 μm2
large and aligned with pixel
centre
Source position 5 cm far along the main Not relevant, 10 cm far along the main
detector axis collimated beam detector axis
Gaussian electronic noise 2.1 keV 0.7 keV 6 keV

Fig. 8. Comparison between simulated and measured spectra of 25 (a) and 50 keV (b) monochromatic synchrotron radiation.

the HEXITEC ASIC and the simulation outputs are shown in Fig. 8. anode of seven 250 μm strips with an intergap of 550 μm (Fig. 9).
Simulation parameter are listed in Table 2. This strip geometry was chosen in order to ensure a high energy
In both cases a remarkable agreement has been achieved with NSD resolution for energy radiation greater than 100 keV without the need
values of 0.012 and 0.018. The greater value of the 50 keV spectrum is for large numbers of readout channels. Each module was connected to
due to the low energy tailing of the experimental peak. This distortion an analog readout system developed by Due2Lab s.r.l. (Parma, Italy).
can be ascribed to scattering effects from the measurement setup that Each readout channel includes CSP (cremat CR110, 140 μs of decay
were not considered in the simulation. The experimental and simulated time), shaper amplifier (2 μs of shaping time) and peak detector.
AFE values are 0.198 and 0.202, respectively, which corresponds to a
A 137 Cs spectrum has been simulated with the same operating
difference of 0.2%.
conditions of the experimental measurement (Table 2). In order to
prevent the detector to collect unwanted scattered photons, the 137 Cs
3.3. Strip detector
radioactive source was collimated with lead and the device was then
placed close to the collimator aperture. The measurement was carried
As last validation we report a comparison with a completely dif-
out at room temperature (see Fig. 10).
ferent detector geometry which was fabricated as part of a flagship
project to develop a radiation detector that was mounted on an Un- The NSD value is 0.0003. The experimental and simulated AC values
manned Aerial Vehicle (UAV). The aim of this project was to develop are 0.815 and 0.824, respectively, which corresponds to a difference
a technology that could aid human operators during the examination of 1%. The experimental and simulated P/C ratio values are 2.44
of radioactively contaminated areas [30,41,42]. The application is and 2.3, respectively, which corresponds to a difference of 5%. In
different from the previous examples: in this case the energy range this case the electronic noise seems to be underestimated in the low
spans up to ∼1 MeV. energy region which leads to an excessive broadening. This is sug-
The prototype detector is a modular system composed of four 20.0 × gested by the discrepancy of the peak-to-valley ratio. At ∼200 keV the
4.5 × 6.0 mm3 CdZnTe detectors with a strip electrode geometry. The backscattering peak can be observed in the measured spectrum which is
detectors were fabricated from standard Redlen spectroscopic grade indeed produced by the physical surroundings of the detector. Since the
CdZnTe. The detectors consisted of a planar cathode and a segmented simulations have been performed assuming a bare crystal, this feature

8
M. Bettelli, N.S. Amadè, D. Calestani et al. Nuclear Inst. and Methods in Physics Research, A 960 (2020) 163663

Fig. 9. 3D model of strip detector, electronics readout system and UAV prototype.

the response function of detectors and allows the calibration and testing
of correction algorithms prior to the actual fabrication of real devices.
In conclusion, the main power of this simulation framework is that
excellent results can be obtained without the need of experimental
measurements, granting large benefits in terms of research and de-
velopment costs. The detector design can be accurately optimized for
specific applications prior to fabrication. Furthermore, information like
the electric field, the weighting field and the induced current transients
can be very useful for the simulation-led design of readout electronics
and correction algorithms [46,47].

Declaration of competing interest

The authors declare that they have no known competing finan-

cial interests or personal relationships that could have appeared to
Fig. 10. Comparison between simulated and measured spectra achieved with stripe
detector and 137 Cs nuclear source.
influence the work reported in this paper.

CRediT authorship contribution statement

does not appear. However, it could be easily implemented in the Geant4
block. Manuele Bettelli: Conceptualization, Methodology, Software, Vali-
dation, Investigation, Writing - review & editing. Nicola Sarzi Amadè:
Conceptualization, Methodology, Validation, Investigation, Writing -
4. Conclusions review & editing. Davide Calestani: Writing - review & editing, Vi-
sualization. Bruno Garavelli: Conceptualization, Writing - review &
A simulation framework based on first principle calculations has editing. Pietro Pozzi: Conceptualization, Writing - review & editing.
been developed and successfully validated by comparison with the Daniele Macera: Conceptualization, Writing - review & editing. Luca
experimental results obtained with real detectors having three different Zanotti: Conceptualization, Writing - review & editing. Carlo Andrea
geometries. The validation has been carried out for a wide range of Gonano: Conceptualization, Writing - review & editing. Matthew C.
photon energies (from 25 keV to 660 keV) and has demonstrated the Veale: Writing - review & editing, Investigation. Andrea Zappettini:
capability of reproducing different radiation–matter interaction mecha- Conceptualization, Methodology, Writing - review & editing, Project
nisms (photoelectric absorption, Compton scattering, CZT fluorescence administration, Supervision.
edges and escape peaks). The simulations proved to correctly replicate
features related to charge carrier trapping mechanisms in CdZnTe References
material that results in the presence of low energy tailing within the
resulting spectra, as demonstrated for the large-pixel detector. The [1] G. Vedrenne, J.-P. Roques, V. Schonfelder, P. Mandrou, G.G. Lichti, A. Von Kien-
simulated spectra show a remarkable agreement with the experimental lin, B. Cordier, S. Schanne, J. Knodlseder, G. Skinner, et al., SPI: The
data which has been quantified with the normalized standard deviation spectrometer aboard INTEGRAL, Astron. Astrophys. 411 (2003) L63–L70.
[2] S. Del Sordo, L. Abbene, E. Caroli, A.M. Mancini, A. Zappettini, P. Ubertini,
(NSD), the area under the Compton edge (AC), the photopeak/Compton
Progress in the development of CdTe and CdZnTe semiconductor radia-
edge ratio (P/C) and the area under the fluorescence and escape peaks tion detectors for astrophysical and medical applications, Sensors 9 (2009)
(AFE). The discrepancies are probably mainly related to secondary 3491–3526.
radiation. In principle, spectral distortions due to backscattering could [3] P. Ubertini, F. Lebrun, G. Di Cocco, A. Bazzano, A.J. Bird, K. Broenstad, A.
be reproduced by properly modelling the physical surroundings of the Goldwurm, G. La Rosa, C. Labanti, P. Laurent, et al., IBIS: The imager on-board
INTEGRAL, Astron. Astrophys. 411 (2003) L131–L139.
detector.
[4] J. Barylak, P. Podgórski, T. Mrozek, A. Barylak, M. Stęślicki, J. Sylwester,
Different mechanisms such as Compton scattering, escape, fluo- D. Ścisłowski, Geant4 simulations of detector response matrix for Caliste-
rescence and annihilation introduce, within the measured spectrum, SO, in: Photonics Applications in Astronomy, Communications, Industry, and
both a continuous background and spurious peaks. Several methods High-Energy Physics Experiments 2014, Vol. 9290, p. 929037.
based on different approaches (stripping method, genetic algorithm, [5] A. Tomal, J.C. Santos, P.R. Costa, A.H. Lopez Gonzales, M.E. Poletti, Monte
Carlo simulation of the response functions of CdTe detectors to be applied in
regularization methods) have been proposed so far [43–45] to remove
x-ray spectroscopy, Appl. Radiat. Isot. 100 (2015) 32–37.
these undesired effects. They all rely on the experimental measurement [6] M. Gerlach, M. Krumrey, L. Cibik, P. Müller, G. Ulm, Comparison of scattering
of the response function of the detector. The accuracy of the presented experiments using synchrotron radiation with Monte Carlo simulations using
simulation framework provides an alternative method for calculating Geant4, Nucl. Instrum. Methods Phys. Res. A 608 (2009) 339–343.

9
M. Bettelli, N.S. Amadè, D. Calestani et al. Nuclear Inst. and Methods in Physics Research, A 960 (2020) 163663

[7] S. Spannagel, K. Wolters, D. Hynds, N.A. Tehrani, M. Benoit, D. Dannheim, N. [27] S. Fatemi, S. Altieri, S. Bortolussi, I. Postuma, G. Benassi, N. Zambelli, M. Bettelli,
Gauvin, A. Nurnberg, P. Schutze, M. Vicente, Allpix2: A modular simulation M. Zanichelli, A. Zappettini, N. Protti, Preliminary characterization of a CdZnTe
framework for silicon detectors, Nucl. Instrum. Methods Phys. Res. A 901 (2018) photon detector for BNCT-SPECT, Nucl. Instrum. Methods Phys. Res. A 903
164–172. (2018) 134–139.
[8] D. Greiffenberg, M. Andra, R. Barten, A. Bergamaschi, M. Bruckner, S. Alvarez, [28] J. Aleotti, G. Micconi, S. Caselli, G. Benassi, N. Zambelli, M. Bettelli, D.
S. Chiriotti, I. Chsherbakov, R. Dinapoli, P. Fajardo, E. Frojdh, et al., Charac- Calestani, A. Zappettini, Haptic teleoperation of UAV equipped with gamma-
terization of GaAs: Cr sensors using the charge-integrating JUNGFRAU readout ray spectrometer for detection and identification of radio-active materials in
chip, J. Instrum. 14–05 (2019) P05020. industrial plants, in: Factories of the Future, Springer, 2019, pp. 197–214.
[9] J. Cammin, S.G. Kappler, T. Weidinger, K. Taguchi, Evaluation of models of [29] T.H. Prettyman, Theoretical framework for mapping pulse shapes in semicon-
spectral distortions in photon-counting detectors for computed tomography, J. ductor radiation detectors, Nucl. Instrum. Methods Phys. Res. A 428 (1999)
Med. Imaging 3 (2016) 023503. 72–80.
[10] J. Cammin, J. Xu, W.C. Barber, J.S. Iwanczyk, N.E. Hartsough, K. Taguchi, A [30] F.P. Doty, H.B. Barber, F.L. Augustine, J.F. Butler, B.A. Apotovsky, E.T. Young,
cascaded model of spectral distortions due to spectral response effects and pulse W. Hamilton, Pixellated CdZnTe detector arrays, Nucl. Instrum. Methods Phys.
pileup effects in a photon-counting x-ray detector for CT, Med. Phys. 41 (2014) Res. A 353 (1994) 356–360.
041905. [31] H.B. Barber, D.G. Marks, B.A. Apotovsky, F.L. Augustine, H.H. Barrett, J.F. Butler,
[11] E.S. Dreier, J. Kehres, M. Khalil, M. Busi, Y. Gu, R. Feidenhans, U.L. Olsen, E.L. Dereniak, F.P. Doty, J.D. Eskin, W.J. Hamilton, et al., Progress in developing
Spectral correction algorithm for multispectral CdTe x-ray detectors, Opt. Eng. focal-plane-multiplexer readout for large CdZnTe arrays for nuclear medicine
57 (2018) 054117. applications, Nucl. Instrum. Methods Phys. Res. A 380 (1996) 262–265.
[12] J.E. Fernandez, V. Scot, L. Sabbatucci, A modeling tool for detector resolution [32] H.H. Barrett, J.D. Eskin, H.B. Barber, Charge transport in arrays of semiconductor
and incomplete charge collection, X-Ray Spectrom. 44 (2015) 177–182. gamma-ray detectors, Phys. Rev. Lett. 75 (1995) 156–159.
[13] M.C. Veale, C. Angelsen, P. Booker, J. Coughlan, M.J. French, A. Hardie, M. Hart, [33] S. Agostinelli, et al., Geant4—a simulation toolkit, Nucl. Instrum. Methods Phys.
J. Lipp, T. Nicholls, A. Schneider, et al., Cadmium zinc telluride pixel detectors Res. A 506 (2003) 250–303.
for high-intensity x-ray imaging at free electron lasers, J. Phys. D: Appl. Phys. [34] G. Ottaviani, Charge carriers transport properties in CdTe measured with time
52–8 (2018) 085106. of flight technique, Rev. Phys. Appl. 12 (2) (1977) 249–254.
[14] Z. He, Review of the Shockley-Ramo theorem and its application in semicon- [35] M. Bettelli, G. Benassi, L. Nasi, N. Zambelli, A. Zappettini, E. Gombia, L. Abbene,
ductor gamma-ray detectors, Nucl. Instrum. Methods Phys. Res. A 463 (2001) F. Principato, D. Calestani, Mechanically stable metal layers for ohmic and
250–267. blocking contacts on CdZnTe detectors by electroless deposition, in: IEEE Nuclear
[15] E. Belas, P. Praus, R.B. James, R. Grill, Š. Uxa, Determination of electric-field Science Symposium and Medical Imaging Conference, 2015, pp. 1–3.
profile in CdTe and CdZnTe detectors using transient-current technique, IEEE [36] G. Benassi, L. Nasi, M. Bettelli, N. Zambelli, D. Calestani, A. Zappettini, Strong
Trans. Nucl. Sci. 59 (2012) 2402–2408. mechanical adhesion of gold electroless contacts on CdZnTe deposited by
[16] J. Fink, H. Krüger, P. Lodomez, N. Wermes, Characterization of charge collection alcoholic solutions, J. Instrum. 12 (2017) P02018.
in CdTe and CZT using the transient current technique, Nucl. Instrum. Methods [37] L. Abbene, N. Zambelli, G. Gerardi, G. Raso, G. Benassi, M. Bettelli, F.
Phys. Res. A 560 (2006) 435–443. Principato, A. Zappettini, High bias voltage CZT detectors for high-flux mea-
[17] A. Zumbiehl, M. Hage-Ali, P. Fougeres, J.M. Koebel, R. Regal, P. Siffert, Electric surements, in: IEEE Nuclear Science Symposium, Medical Imaging Conference
field distribution in CdTe and Cd1- xZnxTe nuclear detectors, J. Cryst. Growth and Room-Temperature Semiconductor Detector Workshop, 2016, pp. 1–6.
197 (3) (1999) 650–654. [38] L. Abbene, G. Gerardi, G. Raso, F. Principato, N. Zambelli, G. Benassi, M. Bettelli,
[18] M. Pavesi, A. Santi, M. Bettelli, A. Zappettini, M. Zanichelli, Electric field A. Zappettini, Development of new CdZnTe detectors for room-temperature
reconstruction and transport parameter evaluation in CZT X-ray detectors, IEEE high-flux radiation measurements, J. Synchrotron Radiat. 24 (2017) 429–438.
Trans. Nucl. Sci. 64 (10) (2017) 2706–2712. [39] B. Thomas, M.C. Veale, M.D. Wilson, P. Seller, A. Schneider, K. Iniewski,
[19] P.N. Luke, Single-polarity charge sensing in ionization detectors using coplanar Characterisation of Redlen high-flux CdZnTe, J. Instrum. 12 (2017) C12045.
electrodes, Appl. Phys. Lett. 65 (1994) 2884–2886. [40] L. Jones, P. Seller, M. Wilson, A. Hardie, HEXITEC ASIC—A pixellated readout
[20] O. Frisch, British atomic energy report, BR-49, 1944. chip for CZT detectors, Nucl. Instrum. Methods Phys. Res. A 604 (2009) 34–37.
[21] A.E. Bolotnikov, N.M. Abdul-Jabbar, S. Babalola, G.S. Camarda, Y. Cui, A. [41] J. Aleotti, G. Micconi, S. Caselli, G. Benassi, N. Zambelli, M. Bettelli, A. Zap-
Hossain, E. Jackson, H. Jackson, J.R. James, A.L. Luryi, et al., Optimization pettini, Detection of nuclear sources by UAV teleoperation using a visuo-haptic
of virtual Frisch-grid CdZnTe detector designs for imaging and spectroscopy augmented reality interface, Sensors 17 (2017) 2234.
of gamma rays, in: Proceedings of SPIE Hard X-ray and Gamma-Ray Detector [42] J. Aleotti, G. Micconi, S. Caselli, G. Benassi, N. Zambelli, D. Calestani, M.
Physics IX, Vol. 6702, 2007, p. 670603-1. Zanichelli, M. Bettelli, A. Zappettini, Unmanned aerial vehicle equipped with
[22] A.E. Bolotnikov, G.S. Camarda, G.A. Carini, G.W. Wright, D.S. McGregor, W. spectroscopic CdZnTe detector for detection and identification of radiological
McNeil, R.B. James, New results from performance studies of Frisch-grid CdZnTe and nuclear material, in: 2015 IEEE Nuclear Science Symposium and Medical
detectors, in: Proceedings of SPIE Hard X-ray and Gamma-Ray Detector Physics Imaging Conference, NSS/MIC, IEEE, 2015.
VI, Vol. 5540, 2004, pp. 33–45. [43] E. Di Castro, R. Pani, R. Pellegrini, C. Bacci, The use of cadmium telluride
[23] P.N. Luke, Unipolar charge sensing with coplanar electrodes-application to detectors for the qualitative analysis of diagnostic x-ray spectra, Phys. Med. Biol.
semiconductor detectors, IEEE Trans. Nucl. Sci. 42 (1995) 207–213. 29 (1984) 1117–1131.
[24] Z. He, Potential distribution within semiconductor detectors using coplanar [44] V. Suman, P.K. Sarkar, Nuclear instruments and methods in physics research a
electrodes, Nucl. Instrum. Methods Phys. Res. A 365 (1995) 572–575. neutron spectrum unfolding using genetic algorithm in a Monte Carlo simulation,
[25] J. Ryan, J.R. Macri, M.L. McConnell, B.K. Dann, M.L. Cherry, T. Guzik, F. Doty, Nucl. Instrum. Methods Phys. Res. A 737 (2014) 76–86.
B.A. Apotovsky, J.F. Butler, Large-area submillimeter resolution CdZnTe strip [45] M. Morháč, V. Matoušek, Complete positive deconvolution of spectrometric data,
detector for astronomy, in: Proceedings of SPIE’s 1995 International Symposium Digit. Signal Process. Rev. J. 19 (2009) 372–392.
on Optical Science, Engineering, and Instrumentation, Vol. 2518, 1995, pp. [46] V. Chaplin, N. Bhat, M.S. Briggs, V. Connaughton, Analytical modeling of pulse-
292–301. pileup distortion using the true pulse shape; applications to Fermi-GBM, Nucl.
[26] C.M. Stahle, A.M. Parsons, L.M. Bartlett, P. Kurczynski, J.F. Krizmanic, L.M. Instrum. Methods Phys. Res. A 717 (2013) 21–36.
Barbier, S.D. Barthelmy, F.B. Birsa, N.A. Gehrels, J.L. Odom, et al., CdZnTe strip [47] K. Taguchi, E.C. Frey, X. Wang, J.S. Iwanczyk, W.C. Barber, An analytical model
detector for arcsecond imaging and spectroscopy, in: Proceedings of SPIE’s 1996 of the effects of pulse pileup on the energy spectrum recorded by energy resolved
International Symposium on Optical Science, Engineering, and Instrumentation, photon counting x-ray detectors, Med. Phys. 37 (2019) 3957–3969.
Vol. 2859, 1996, pp. 74–84.

10