Download as pdf or txt
Download as pdf or txt
You are on page 1of 10

IEEE TRANSACTIONS ON NUCLEAR SCIENCE, VOL. 52, NO.

3, JUNE 2005 589

Factors Limiting the Performance


of CdZnTe Detectors
A. E. Bolotnikov, Member, IEEE, G. C. Camarda, G. W. Wright, and R. B. James, Fellow, IEEE

Abstract—In the past few years, significant progress has been temperature, introduces additional obstacles in the implementa-
achieved in the development of room-temperature semiconductor tion of actual CZT spectroscopic devices.
detectors, particularly those based on CdZnTe (CZT) crystals. Three major parameters characterize the performance of a
Several types of electron-transport-only detectors have been de-
veloped: pixel, coplanar-grid, cross-strip, drift-strip, orthogonal radiation detector: the detection efficiency, the size of the ac-
coplanar strip, and virtual Frisch grid, many of which are now tive area, and the energy resolution. Both the efficiency and the
commercially available. Despite all these varieties in the detector size of the active area are determined by the size of the avail-
designs, they have many common features and problems. This able CZT crystals, which are currently small in comparison to
review summarizes the common detector design constraints and
related factors limiting performance of CZT detectors: bulk and HPGe detectors, less than cm . The ultimate solution of
surface leakage currents, surface effects, properties of Schottky this problem lies in a way of improving CZT crystal growth
contacts and surface interfacial layers, charge sharing and loss techniques. Meanwhile, arrays or stacks of a large number of
in multielectrode devices, charge transport nonuniformities, and relatively small CZT detectors can be used to build large area
fluctuations in the pulse height for long-drift-length devices. We
high-efficiency detection systems.
also describe unique capabilities at Brookhaven National Labo-
ratory, Upton, NY, for CZT device characterization and recent Even with the small size of CZT crystals, it takes some ef-
progress utilizing these tools. fort to produce high-performance CZT detectors. In general,
Index Terms—CdZnTe (CZT), gamma-ray detectors, semicon-
the energy resolution of a semiconductor radiation detector is
ductor radiation detectors. determined by three terms: 1) fluctuation in the number of elec-
tron-hole pairs produced, 2) fluctuation in the number of elec-
trons and holes collected, and 3) electronic noise. The first term
I. INTRODUCTION depends on the mean energy for electron-hole pair generation,
, and the Fano factor, , where is the par-
S EMI-INSULATING CdZnTe (CZT) has been introduced
to overcome the problems associated with high-purity Ge
(HPGe) (requires cooling) and Si(Li) (low stopping power) with
ticle’s energy ( denotes the variance of the total number of
electron-hole pairs produced). In high-quality crystals, the Fano
potential spectroscopic performance that approaches HPGe and factor is . The second term accounts for fluctuations of
Si(Li) radiation detectors [1]. the trapped charge. In the case of HPGe detectors—which can
CZT detectors have higher stopping power and operate at be considered as the ideal case—the trapping centers are dis-
room temperature, but their active volume and spectral response tributed uniformly. Thus, the contribution due to the fluctuation
are still behind Ge detectors. The primary problem lies in phys- of the trapped charge (variance) is , where
ical and electronic properties of currently available CZT crys- and are the carriers drift and life times. (If ,
tals. The maximum crystal size and quality of HPGe are still then this term can be neglected.) In contrast to HPGe, the traps
superior over CZT. The sensitive volume of HPGe detectors is distribution is not uniform in CZT material and, as a result, the
typically of hundreds of cm . The volume of good quality CZT above equation may be not valid. Indeed, the charge generated
crystals is limited to a few cm . CZT is comparably “slow” ma- by an ionizing particle is localized inside a small area of the de-
terial: electron and hole drift mobilities in CZT are and tector, and if the trap distribution is not uniform, the collected
cm /Vs, respectively, while the same parameters in HPGe charge could have additional fluctuations. The third term is the
are 3900 and 1900 cm /Vs. Correspondingly, -products are electronic noise, which is typically dominated by the shot noise
cm /V (electrons) and cm /V (holes) for due to detector leakage current. In the case of the CZT detectors
CZT and cm /V (both electrons and holes) in HPGe. These operating at room temperature, the leakage current becomes the
differences in parameters lead to the incomplete charge collec- most critical parameter determining the energy resolution for
tion in CZT devices, particularly for holes. Furthermore, the devices with good charge collection.
mean energy for electron-hole generation in HPGe is eV, Usually, CZT detectors operate as electron-only devices to
in comparison to eV in CZT, which results in smaller output eliminate the influence of the uncollected holes. In such de-
signals for the CZT detectors. It is clear from the above com- vices, the electron contribution into the output signal is pro-
parisons that CZT material, even though it can be used at room portional to the total charge produced by an ionized particle.
As for the holes, their contribution depends on the detector ge-
ometry and where electron-hole pairs were originally produced.
Manuscript received July 9, 2004. As a rule, the uncollected holes degrade the energy resolution
The authors are with the Brookhaven National Laboratory, Upton, NY 11793
USA (e-mail: bolotnik@bnl.gov). of CZT gamma-ray detectors; this effect is especially strong
Digital Object Identifier 10.1109/TNS.2005.851419 in planar devices with monolithic contacts. Several contact de-
0018-9499/$20.00 © 2005 IEEE

Authorized licensed use limited to: Indian Institute of TechnologyDharwad. Downloaded on March 15,2022 at 07:08:16 UTC from IEEE Xplore. Restrictions apply.
590 IEEE TRANSACTIONS ON NUCLEAR SCIENCE, VOL. 52, NO. 3, JUNE 2005

signs have been proposed to reduce the effect of the uncollected the grid was kept at zero potential. In the case of surface current,
holes in CZT material [2]–[9]. Two of the designs, pixel and the voltage was applied to the grid while the cathode was kept
coplanar-grid, are widely used in CZT gamma-ray detectors. All at some potential.
these designs have an analog among gaseous detectors—an ion-
ization chamber with a shielding (Frisch) grid. The only differ- B. Bulk Leakage Current
ence is that there is no actual shielding grid in CZT devices: the Fig. 1 shows typical – curves measured for three CZT
grid is virtually created by choosing a certain electrode configu- samples. The first sample (with sputtered Pt contacts) was
ration. As in the case of gaseous detectors [10], the virtual grid in grown and processed by eV-Products, Inc., the second one
CZT devices does not provide 100% “shielding” from the holes. (also with sputtered Pt-contacts) was grown by Imarad and
As a result, the peak in the response function of the device is far processed by GSFC, and the third one (with electroless Au
from the -function: it has a finite geometrical width that adds contacts) was grown by Yinnel Tech., Inc., and processed at
to the total energy resolution. Brookhaven National Laboratory. The – curves for the first
In this paper, we review the major factors limiting perfor- two samples were taken from [18].
mance of current CZT detectors. It was shown in [17] that the bulk current could be interpreted
by considering a CZT detector as a metal–semiconductor–metal
II. EFFECTS OF SURFACE AND BULK LEAKAGE CURRENTS system (MSM) with two back-to-back Schottky barriers at the
Bulk and surface conductivity are two important parameters contacts (see Appendix). Thus, the differences in the curves in
that 1) determine the total leakage current and 2) affect the Fig. 1 reflect the differences in material bulk resistivity and sur-
charge collection efficiency in CZT devices, e.g., for pixel and face treatments of the CZT samples. Note that, among the three
coplanar-grid detectors. The latter effect can be explained as fol- curves, only the – curve measured from the Imarad sample
lows. In the absence of surface conductivity, the electric field actually resembles a diode-like dependence because of its lower
lines are expected to never intersect the nonmetallized surfaces bulk resistivity and higher voltage bias required to deplete the
because of the surface charge built up in the areas between the crystal.
contacts. However, in real CZT detectors, the surface is conduc- Four distinguished regions can be seen in – curves. At
tive. As a result, the built-up charge leaks away and field lines the low biases, V, a voltage drop across a Schottky
can intersect the surface areas between the contacts where the barrier is very small, and the current is limited by the bulk
signal charge gets trapped [11]–[13]. To overcome this effect a resistivity of CZT material (Ohm’s law). This region can be
differential bias is applied between the adjacent electrodes (e.g., used to estimate the bulk resistivity, e.g., for the samples
in coplanar-grid devices), so a strong electric field inside the gap represented in Fig. 1, one can find , and
forces the electrons on the surface to move toward one of the Ohm-cm for Imarad, eV-Products, and Yinnel Tech.
contacts, significantly reducing the charge loss. In the case of samples, respectively.
pixel detectors, thin steering electrodes (a grid) are placed be- At higher biases, V, the voltage drop across the Schottky
tween the pixel contacts [14], [15] to eliminate the charge loss. barrier starts gradually increasing while the undepleted (resis-
The main drawback of using the steering electrodes (or ap- tive) layer of the sample gets thinner and thinner. This transition
plying a differential bias) is the extra surface leakage current is reflected as a sublinear dependence in the – curves (region
component that significantly increases the total leakage current 2), e.g., in the case of the Imarad sample. Since the
and thermal noise. Moreover, as reported in [16], the surface charge transport across the interface is limited by carrier dif-
conductivity strongly depends on the cathode bias. Thus, the fusion in the space-charge region of CZT, the saturation current
total leakage current turns out to be more than the sum of two follows the voltage dependence of the electric field strength near
separately measured bulk and surface components. the metal–semiconductor interface [19].
Another important feature related to charge collection effi- For the fully depleted material, the electric field near the
ciency is the nonuniformity of the electric field over the detector contact is a linear function of the applied voltage. From this
thickness caused by the Schottky-barrier contacts. In the over- moment, the – curve follows a linear dependence again
simplified case of uniform distribution of space charge, the elec- resembling ohmic behavior but with very high effective bulk
tric field strength decreases linearly toward the anode (for n-type resistivity (region 3). This region, however, is not always seen
material). In reality, because of the deep trapping levels in CZT, in the – curves because the current may start rising expo-
the field seems to decrease more rapidly, especially at low tem- nentially due to the potential barrier lowering effect (region 4).
peratures. The decrease of potential barrier is caused by the voltage de-
veloped across a thin interfacial layer existing between the metal
A. Experimental Setup and the bulk of CZT. Usually, this layer is formed as a side ef-
Reference [17] describes the experimental setup and mea- fect during the fabrication process, e.g., if the detector was ex-
surement procedure that allows one to separate bulk and sur- posed to air before depositing contacts. The exact properties of
face leakage current components and to achieve a steady-state this layer are unknown: it could be a dielectric, e.g., TeO , or
current and temperature stable conditions during the measure- a p-type semiconductor, e.g., Te, which forms a p–n junction
ments. All results described here were obtained with pixel detec- with CZT. Since the layer is thin, the electrons entering from
tors whose anode patterns had square pixel contacts surrounded the metal can tunnel across the layer with some transmission ef-
by guard strips (a grid) [14]. In the case of bulk current mea- ficiency. Thus, regardless of the precise nature of the interfacial
surements, the voltage source was applied to the cathode, while layer, its effect on the leakage current can be described with help

Authorized licensed use limited to: Indian Institute of TechnologyDharwad. Downloaded on March 15,2022 at 07:08:16 UTC from IEEE Xplore. Restrictions apply.
BOLOTNIKOV et al.: FACTORS LIMITING THE PERFORMANCE OF CdZnTe DETECTORS 591

Fig. 1. Bulk I –V curves measured with CZT samples fabricated by different vendors.

of two parameters: an empirical transmission coefficient, which C. Surface Leakage Current Measurements
takes into account the reflection of the electron, and a potential
The surface conductance becomes an important factor for
lowering, which is in the first approximation proportional to the
contact patterns with steering (biased) electrodes, e.g., in
applied bias.
coplanar-grid or pixel devices. It is also important for the vir-
The solid lines in Fig. 1 represent the results of the – curve tual Frisch-grid detectors, since the side surface leakage current
fittings with the MSM model as described in [17]. For the – is expected to ensure a uniform distribution of the electric field
curves shown in Fig. 1, the fit gives similar values for the poten- inside the detector. In such devices, the surface leakage current
tial barrier heights: 0.740 eV (Imarad) and 0.784 eV (eV-Prod- becomes the dominant component to the total leakage current.
ucts), and 0.810 eV (Yinnel Tech.), but different values for the The surface conductance in CZT detectors is apparently due
voltage required to deplete the samples: V (eV-Products to the presence of a low-resistivity surface layer, which has a
and Yinnel Tech.) and much higher bias V (Imarad). The different stoichiometry and defect structure, in comparison with
fact that the Imarad detector requires much higher voltage to the bulk-CZT band structure. It is well known that a damaged
deplete the crystal is due to the lower bulk resistivity of this surface layer of nonstoichiometric material is formed on the sur-
material. In [20], the barrier height of Au–CZT Schottky con- face of semiconductors after the dicing and polishing. This layer
tact fabricated by eV-Products was measured using a different is normally removed by chemical etching; however, in the case
technique—the photocurrent method. The reported value was of CZT, the bromine enchant usually leaves a Te-enriched CZT
eV, which is very close to that estimated from the surface [21].
– curve shown in Fig. 1. Fig. 2 shows the surface – curves measured for a typical
An important conclusion that can be made from Fig. 1 is that eV-Products CZT detector at zero bias at the cathode [14]. As
the bulk leakage current is not determined by the bulk resistivity shown, in the beginning the current rises as a linear function of
of CZT material, but rather by the properties of the contacts (i.e., voltage applied across the gap between the pixel contact and the
by the contact fabrication process and possibly the choice of grid (see Section II-A). Above a certain transition voltage, the
electrode material). current starts rising as a power law function, , with

Authorized licensed use limited to: Indian Institute of TechnologyDharwad. Downloaded on March 15,2022 at 07:08:16 UTC from IEEE Xplore. Restrictions apply.
592 IEEE TRANSACTIONS ON NUCLEAR SCIENCE, VOL. 52, NO. 3, JUNE 2005

and is an analog to an active channel in a field-effect transistor.


The changes in surface leakage shown in Figs. 3 and 4 can be at-
tributed to the depletion (or enhancement) of this highly doped
surface layer when bias is applied between two contacts on the
surface (MSM structure) or to the cathode (field effect). For ex-
ample, in Fig. 4, the – curve measured at positive 300 V
corresponds to the fully depleted layer, while the negative bias
enhances the channel up to its maximum width. The fact that
at high biases all three curves converge means that the surface
layer can be depleted by applying voltage either on the cathode
or between the pixel contact and the grid. As shown in Fig. 4, the
current increases at negative voltages and decreases at positive
biases, i.e., the detector resembles a p-type channel field-effect
transistor operating in depletion mode.
The same MSM model (originally proposed to fit bulk cur-
rents) was applied to fit the surface leakage current – curves
measured at the negative cathode bias that enhances the surface
conductivity [18]. For example, the fit results are shown in Fig. 3
as solid lines for two cases: 300 and 0 V on the cathode. The fit
Fig. 2. Surface current I –V curve (large scale measurements) from [14].
yields the following parameters: a potential barrier height of 0.51
eV, a carrier mobility of 100 cm /Vs, and a specific resistivity
of Ohm-cm, with a grid voltage required to deplete the
. The transition voltage (or voltage of crossover from Ohm’s surface layer of 2.5 V. These parameters depend on the assumed
law to power law) is related to a high-injection condition at the thickness of the layer and the effective Richardson constant,
contact, but the current does not rise exponentially because the which were assumed to be 50 nm and 12 A/cm K , respectively.
space charge accumulated between the contacts limits current In the case when no interfacial layer is considered for the
injection from the contacts (see, e.g., [22]). surface, we applied the same fit to the surface current data ob-
More accurate – measurements, however, revealed several tained from the eV-Products sample, which gave the following
features similar to those seen in the bulk – curves. Fig. 3 results: a potential barrier height of 0.53 eV, a carrier mobility
shows the – curve of the surface leakage current measured of 100 cm /Vs, a specific resistivity of Ohm-cm, with
between a pixel contact and the grid at three representative a grid voltage required to deplete the channel surface layer of
cathode biases, 300, 0, and 300 V. In this example, the data 0.6 V. These results also point out that the surface layer exhibits
were taken with the CZT sample grown by Imarad, Inc., and the properties of low resistivity p-type material.
processed at NASA/GSFC [18]. As shown, the curves measured It is interesting to compare the potential barrier heights es-
at 300 V cathode bias have three distinct regions: linear below timated from the – and – measurements, where is the
0.1 V, sublinear around 1 V, and linear again between 10 and temperature. In [18], the temperature dependence of the bulk
50 V (at higher biases, the current starts rising exponentially and surface leakage currents (eV-Products sample) were mea-
and eventually becomes space charge limited as shown in sured and fitted with a function , which
Fig. 2). Such an “s”-shape of the – curve is typical (e.g., represents the temperature dependence of the diffusion-limited
Fig. 1) of a semiconductor device with two Schottky-barrier saturation current across the Schottky barrier. Here, is the po-
contacts. The “s”-shaped feature is less pronounced at 0 V on tential barrier height. The estimates for were found to be 0.57
the cathode, and at 300 V, the curve is almost linear in a whole and 0.84 eV for the surface and the bulk currents, respectively.
range of voltages. Despite differences at low biases, all of the As shown, these numbers are different from those obtained from
measured – curves converge at high biases. the – measurements. We believe that the barrier height of
Fig. 4 shows the variations of the surface current versus 0.51 eV (obtained for the surface current from – measure-
cathode bias measured with the same detector at 0.05 V between ments) is underestimated because of the two unknown param-
the pixel contact and the grid (this bias corresponds to the ohmic eters assumed in [18]. On the contrary, the height of 0.84 eV
region of the surface current, see Fig. 3), [18]. As shown, the (obtained for the bulk current from – measurements) is over-
measured dependence reflects two levels of surface conductance estimated because the – curve also reflects a temperature de-
over a narrow voltage range where the surface resistivity is pendence of the electron flux tunneling through the interfacial
switched from a low level to a higher level. Similar behavior layer.
was observed for the samples fabricated by eV-Products [18]. Based on previous discussions, one can conclude that p-type
layers on the surface result in the p-i-p band structure of the
D. Field Effect CZT detector shown in Fig. 5. The actual barrier height seen by
Based on the work by Rouse et al. [21], the surface of the electrons entering CZT from the metal is higher by an amount
CZT detector consists of an oxide (dielectric) layer and Te (or of 0.29 eV due to a p–n junction at the surface. By comparison,
Cd)-enriched layer that can be treated as a highly doped surface the increased value in the potential barrier height reported by
layer of CZT. This layer forms a p–n junction with the CZT bulk Morton et al. [20] is 0.2 eV.

Authorized licensed use limited to: Indian Institute of TechnologyDharwad. Downloaded on March 15,2022 at 07:08:16 UTC from IEEE Xplore. Restrictions apply.
BOLOTNIKOV et al.: FACTORS LIMITING THE PERFORMANCE OF CdZnTe DETECTORS 593

Fig. 3. Surface current I –V curve (small scale measurements) from [18].

Fig. 4. Surface leakage current (resistivity) versus cathode bias from [18].

Authorized licensed use limited to: Indian Institute of TechnologyDharwad. Downloaded on March 15,2022 at 07:08:16 UTC from IEEE Xplore. Restrictions apply.
594 IEEE TRANSACTIONS ON NUCLEAR SCIENCE, VOL. 52, NO. 3, JUNE 2005

IV. FLUCTUATIONS IN CHARGE LOSSES


The main problem related with CZT material is the poor hole
collection efficiency. Several contact designs have been pro-
posed to reduce the effect of the uncollected holes; three of
them—pixel, coplanar-grid, and drift-strip—are used in large-
volume CZT devices. Despite the fact that these designs are very
effective in shielding the holes, the theoretical limit of the en-
ergy resolution in CZT detectors (the statistical limit) has never
been achieved in the large-volume CZT devices. The typical res-
olution obtained with thick, 5 mm, CZT detectors is 2%–3% at
662 keV. This contradicts the fact that the statistical limit was in-
Fig. 5. Band structure of the metal–CZT interface. deed achieved with thin, less than 2-mm thick, and small area (or
pixel) devices. In other words, the energy resolution degrades in
III. CHARGE SHARING EFFECT those devices in which electrons drift long distances ( 5 mm)
toward the anode electrode (long-drift devices).
Charge sharing events between adjacent contacts is another
important factor limiting the performance of multielectrode Originally, the energy degradation in long-drift devices was
CZT devices such as pixel and coplanar-grid detectors. The attributed to the two effects (apart from obvious structural
mean squared width of the charge cloud after drifting a distance defects in crystals such as grain boundaries, pipes, inclusions,
under the influence of the electric field is given by one-di- etc.): large-scale gradual ( 1 mm) variations in the CZT
mensional diffusion as . material properties ( -product), and the dependence of the
Here, is the drift time, is the diffusion coefficient, is the amount of trapped electrons (electron carrier lifetime) versus
electric field strength, and V at room tempera- electron drifting time. These effects can be minimized by
ture. For a 2-mm-thick detector with an electric field of 1500 using multielectrode contact patterns with a depth correction
V/cm, equals to 24 m [or 56 m full-width at half-maximum technique. However, even with these improvements, the mea-
(FWHM)]. This example shows that for small contact sizes sured energy resolution in thick devices was found to remain
the total charge produced by an incident particle can be shared above its statistical limit. This suggests there could be some
between two or more adjacent contacts. Thus, to recover the other unaccounted effects degrading the energy resolution in
total charge one must add the signals read out from two or more long-drift CZT devices [26]–[28].
contacts, which increases the electronic noise and complexity It is clear that the degradation in the energy resolution is cor-
of the readout system. Moreover, since the electric field lines related with large drifting distances for electrons inside the CZT
intersect the surface areas between the contacts, some fraction detectors. Thus, it is reasonable to assume that the cause of the
of the charge could be lost in the gaps between the contacts. In energy resolution degradation is the fluctuations in the charge
turn, the fluctuations in these charge losses degrade the energy loss along the electron paths. An exact mechanism of these fluc-
resolution of the device. tuations is unknown. It can be shown, however, that these fluctu-
The easiest way to illustrate the charge sharing and charge ations are not related to the electron loss due to trapping by im-
loss between the contacts is to plot the correlation of the sig- purities. Indeed, the absorption of a 662-keV photon produces
nals, and in Fig. 6, from two adjacent electrodes for the electron-hole pairs in CZT crystal, of which 10%
events taking place between the contact boundaries [25]. Each are usually lost. If one assumes that trapping centers are ran-
event is represented by a single dot; the dots are located around domly distributed inside the material and each center traps a
the curved line (for events with the same absorbed energies) as single electron, then the standard deviation of the number of
shown in Fig. 6. If all charge is collected, one can expect to the lots electrons is simply the square root of 10% of the total
see a straight line, since , where is the number of the electrons produced, . From here, the
total injected energy (a collimated laser beam was used to mea- relative fluctuation is less than 0.1%. This example shows that
sure curves shown in Fig. 6). The bending indicates that there is the uniformly distributed trapping centers cannot explain large
charge loss in the area between the contacts near the surface. fluctuations in the charge loss.
The straightforward solution of this problem is to make very The degradation of energy resolution can be explained if we
small gaps between the electrodes [23], but it is not always ac- assume that the CZT material contains randomly distributed mi-
ceptable, e.g., in pixel detectors, since the contact areas and as- croscopic areas (centers) with very high local concentrations of
sociated load capacitors become too large. Another solution is to traps. The fluctuations in number of such centers encountered
place a thin steering strip [14] or several strips [24] in the gaps by the electron cloud as it drifts toward the anode and the fluc-
between the contacts. A certain differential bias is required to tuations in the charge loss in each of these centers may account
push electric field lines off the steering electrodes. The mag- for the observed fluctuations in the total collected charge.
nitude of the bias depends on both the width of the gap and If one assumes that the total number of the electrons lost due
steering electrode, but even at very high bias the electrons can to trapping is , and each trapping center can only trap a single
still reach the surface (due to diffusion) and become trapped. electron, then the dispersion of the number of the lost charge
This effect adds fluctuations to the total collected charge and simply equals . However, if one assumes
degrades the energy resolution of the detector. that the charge is trapped by a fewer number of trapping centers

Authorized licensed use limited to: Indian Institute of TechnologyDharwad. Downloaded on March 15,2022 at 07:08:16 UTC from IEEE Xplore. Restrictions apply.
BOLOTNIKOV et al.: FACTORS LIMITING THE PERFORMANCE OF CdZnTe DETECTORS 595

Fig. 6. Correlation of the signals readout from two adjacent pixels from [25].

but with high trap concentration, then one can write for the total The origin of these centers can be related to microscopic de-
number of the lost electrons , where is the average fects existing inside CZT material and by the inherent nature
number of electrons trapped by a single center, and is the of these microscopic defects to accumulate a large number of
average number of centers encountered by the electron cloud traps. In the case of CZT, e.g., such defects could be inclusions
as it drifts toward the anode. If , where and precipitates.
and are the electron drift and life times, then, as a first order of This mechanism could also explain the poor energy resolution
approximation, and can be treated as independent variables. observed with long-drift CdZnTe detectors described in the liter-
Using the error propagation formula, one can find ature. In such cases, the depth correction technique only helps to
eliminate the fluctuations in the collected charge due to random
(1) distribution of the interaction points. New results were recently
reported by Amman et al. [28] who studied the effects of elec-
where and are dispersions of and , respectively. tron trapping nonuniformity in CZT detectors with alpha parti-
As in the previous case, one can put . If we also cles. However, due to the experimental setup limitations, they
assume that , then could not separate effects caused by large-scale nonuniformity
of CZT material and the microscopic defects described above.
(2)
V. CONCLUSION
If then the second term in (2) can be neglected. From Three major effects limiting the performance and ultimately
here the energy resolution in CZT devices were discussed: bulk and
surface leakage currents, effects related to the charge sharing
FWHM (3) between the electrodes, and fluctuations in charge losses along
the electron cloud travel path. In CZT devices with Au and Pt
This means that the fluctuations of the trapped charge are di- contacts, the total leakage current is limited by the properties
rectly proportional to the distance traveled by the electron cloud. of the Schottky barriers at the metal–semiconductor interfaces.

Authorized licensed use limited to: Indian Institute of TechnologyDharwad. Downloaded on March 15,2022 at 07:08:16 UTC from IEEE Xplore. Restrictions apply.
596 IEEE TRANSACTIONS ON NUCLEAR SCIENCE, VOL. 52, NO. 3, JUNE 2005

The surface conductance affects the electric-field lines distribu- [30] considered two additional effects: the electron transmis-
tion, which results in charge losses between the adjacent anode sion across the interfacial layer and the lowering of the potential
contacts in multielectrode devices. Fluctuations in charge losses barrier height caused by the voltage drop across the interfacial
due to electron trapping as the electron cloud drifts toward the layer. The first effect leads to the reduction of the thermionic
anode could be an additional factor degrading the energy reso- flux of the electrons. This is described by means of the trans-
lution in long-drift CZT devices. mission coefficient . The second effect is a weak dependence
of the barrier height on the applied voltage. In the first approxi-
APPENDIX mation, this dependence is described as a linear function of the
applied bias:
A commonly accepted model of the Schottky barrier in
metal–semiconductor systems is the so-called thermionic-dif- (A5)
fusion theory developed by Crowell and Sze [29]. It gives a
general formula for the saturation current of the reverse where is the potential barrier at zero bias, and is a model
biased device that can be evaluated numerically. There are parameter. This approach can be justified theoretically in the
two analytical limits of the exact theory historically known case of a thin insulating layer (e.g., oxide layer) between the
as thermionic and diffusion approximations which lead to the metal and semiconductor [31].
same diode-like – dependence, as follows: One more feature should be mentioned when the above theory
is applied to a metal–CZT–metal system. Usually, it is assumed
(A1) that the undepleted part of the semiconductor sample has a neg-
ligibly low resistance. Hence, the entire voltage drops across the
but give different bias dependences for the saturation currents. depleted layer and the electric field strength—a key parameter in
An important assumption made in the diffusion approximation the theory—can be easily evaluated based on the Schottky-de-
is that the free carriers near the contact are considered to be pletion-layer model. In this case, the width of the depletion layer
at thermal equilibrium with the electrons in the metal contact, depends as a root square of the bias, and so does the electric field
which means that the carrier concentration near the metal–semi- at the reversed barrier side of the sample. (This is an important
conductor interface, , is independent of the applied voltage feature that allows us to identify the type of contact behavior
and can be related to the carrier concentration in the bulk, , in the diffusion-limited case.) In the case of CZT, the resistance
as of the undepleted bulk cannot be neglected, and as a result, the
dependence of the saturation current, the electric field, and the
(A2) width of the depletion layer cannot be calculated independently.
All these dependencies become part of the system of equations
Here, is the built-in potential. Thus, in the diffusion limit, derived based on the current continuity principle.
the saturation current is simply given by In the Schottky-depleted-layer approximation, if a negative
voltage, , is applied to the cathode, the electric
(A3) field distribution, , inside both the depleted and undepleted
regions of the detector can be written as
where and are the electron mobility and the electric-field
strength near the contact, respectively. This is true in so-called
“slow” semiconductors; however, if the electron drift velocity (A6)
becomes fast enough to break the equilibrium near the contact
and
(e.g., at high biases) the diffusion approximation is no longer
valid. In the extreme case (fast semiconductors) when the car- (A7)
riers near the interface are rapidly swept away in either direc-
tion depending on the polarity of the applied bias, the saturation In the above equations, is the width of the depleted layer,
current is primary limited by the flux of the carriers thermioni- is the thickness of the CZT crystal, and is the electric field
cally emitted over the potential barrier. These conditions corre- strength inside the undepleted bulk (same as at the anode),
spond to the thermionic emission approximation that gives the is the permittivity of CZT, is the electron charge, is the
following expression for the saturation current: concentration of ionized donor centers, and is
the potential at the edge of the depleted layer
(A4) . Using the boundary condition at the cathode and at the
edge of the depleted layer, the width of the depleted layer can
The parameter is called the effective Richardson constant, be found from the following equation:
(in A-cm K ), where and are the
effective and free electron masses. is the temperature, and (A8)
is the barrier height.
The next step in the generalization of the Crowell and Sze where is the built-in voltage or diffusion potential at the
theory is to include the effects related to the interface layer cathode. From this equation, can be calculated if or is
normally existing between the metal and semiconductor. Wu known. If and then

Authorized licensed use limited to: Indian Institute of TechnologyDharwad. Downloaded on March 15,2022 at 07:08:16 UTC from IEEE Xplore. Restrictions apply.
BOLOTNIKOV et al.: FACTORS LIMITING THE PERFORMANCE OF CdZnTe DETECTORS 597

, and – , i.e., at the small ap- where is the equilibrium free electron concentration (we as-
plied biases the current follows Ohm’s law. The reach-through sume that CZT is an n-type semiconductor). Substituting (A17)
voltage —the voltage required to deplete the whole CZT into (A13) and using from (A8) and (A11), and from
sample—is given by (A12) one can numerically calculate the – dependence.

(A9)
REFERENCES
where is the electric field strength at the anode when the [1] T. E. Schlesinger and R. B. James, “Semiconductors and semimetals,”
in Semiconductors for Room Temperature Nuclear Detector Applica-
cathode is at , i.e., . Notice, that when
tions. San Diego, CA: Academic, 1995, vol. 43.
the bulk resistance is neglected, , and (A9) becomes [2] P. N. Luke, “Unipolar charge sensing with coplanar electrodes,” IEEE
the standard expression for the flat-band voltage—a parameter Trans. Nucl. Sci., vol. 42, no. 4, pp. 207–213, Aug. 1995.
usually defined to characterize the back-to-back barrier system. [3] H. H. Barrett, J. D. Eskin, and H. B. Barber, “Charge transport in arrays
of semiconductor gamma-ray detectors,” Phys. Rev. Lett., vol. 75, no. 1,
For the voltages above p. 156, 1995.
[4] K. Parnham, C. Szeles, K. Lynn, and R. Tjossem, “Performance im-
(A10) provement of CdZnTe detectors using modified two-terminal electrode
geometry,” in Proc. SPIE Conf. Hard X-Ray, Gamma Ray Neutron De-
tector Physics, vol. 3768, 1999, pp. 49–54.
Correspondingly, the electric-field strength at the cathode is [5] G. Montemont, M. Arques, L. Verger, and J. Rustique, “A capacitive
given by Frisch grid structure for CdZnTe detectors,” IEEE Trans. Nucl. Sci., vol.
48, no. 3, pp. 278–281, Jun. 2001.
[6] V. Gostilo, C. Budtz-Jorgensen, I. Kuvvetli, D. Gryaznov, I. Lisjutin,
(A11) and A. Loupinov, “The development of drift-strip detectors based on
and CdZnTe,” IEEE Trans. Nucl. Sci., vol. 49, no. 5, pp. 2530–2534, Oct.
2002.
(A12) 2 2
[7] A. Shor, I. Mardor, and Y. Eisen, “Performance of 1 1 1 cm pixel
CdZnTe gamma detectors,” IEEE Trans. Nucl. Sci., vol. 49, no. 4, pp.
1935–1940, Aug. 2002.
According to Wu [30], the reverse current density, (A/cm ), [8] D. S. McGregor and R. A. Rojeski, “Performance of CdZnTe geomet-
over the barrier at the cathode is expressed as rically weighted semiconductor Frisch grid radiation detectors,” IEEE
Trans. Nucl. Sci., vol. 46, no. 3, pp. 250–259, Jun. 1999.
[9] B. A. Brunett, J. M. Van Scyoc, T. E. Schlesinger, and R. B. James,
“The spectral response of CdZnTe gamma-ray detectors as measured by
(A13) gamma-ray mapping,” Nucl. Instrum. Methods Phys. Res. A, vol. 458,
pp. 76–84, 2001.
[10] I. O. Bunemann, T. Cranshaw, and J. Harvey, “Design of grid ionization
where is the effective Richardson constant, is the tempera- chambers,” Can. J. Res., vol. A27, pp. 191–206, 1949.
ture, is the thermal velocity, is the transmission coefficient [11] A. E. Bolotnikov, W. R. Cook, F. A. Harrison, A. S. Wong, S. M.
through the oxide layer, the series resistance of the bulk, and Schindler, and A. C. Eichelberger, “Charge loss between contacts of
CdZnTe pixel detectors,” Nucl. Instrum. Methods Phys. Res. A, vol.
. is an effective diffusion velocity that can be 432, pp. 326–331, 1999.
calculated analytically if (A6) is used to approximate the field [12] T. H. Prettyman, M. K. Smith, P. N. Luke, M. Amman, and J. Lee, “In-
distribution in the depleted layer. In this case, is simply the vestigation of charge collection in multi-electrode CdZnTe detectors,”
in Proc. SPIE, vol. 3768, 1999, pp. 27–36.
electron drift velocity at the cathode, namely
[13] M. Amman and P. N. Luke, “Position-sensitive germanium detectors for
gamma-ray imaging and spectroscopy,” in Proc. SPIE, vol. 4141, 2000,
(A14) pp. 144–156.
[14] A. E. Bolotnikov, S. E. Boggs, C. M. Hubert Chen, W. R. Cook, F. A.
Harrison, and S. M. Schindler, “Investigation of optimal contact geome-
where is the electron mobility ( 1000 cm /Vs). The effec- tries for CdZnTe pixel detectors,” in Proc. SPIE, vol. 4141, 2000, pp.
tive Richardson constant is related to the thermal velocity by 243–252.
[15] E. Kalemci and J. L. Matteson, “Charge splitting among anodes of a
CdZnTe strip detector,” in Proc. SPIE, vol. 4141, 2000, pp. 235–242.
[16] A. E. Bolotnikov, C. M. Hubert Chen, W. R. Cook, F. A. Harrison, I.
(A15) Kuvvetli, and S. M. Schindler, “Effects of bulk and surface conductivity
on the performance of CdZnTe pixel detectors,” IEEE Trans. Nucl. Sci.,
where is the effective density of the states in the conduction vol. 49, no. 4, pp. 1941–1949, Aug. 2002.
band given by [17] A. E. Bolotnikov, S. E. Boggs, C. M. Hubert Chen, W. R. Cook, F. A.
Harrison, and S. M. Schindler, “Properties of Schottky type contacts on
high-resistivity CdZnTe detectors,” Nucl. Instrum. Methods Phys. Res.
(A16) A, vol. 482, pp. 395–407, 2002.
[18] A. E. Bolotnikov, C. M. Hubert Chen, W. R. Cook, F. A. Harrison, I.
Kuvvetli, S. M. Schindler, C. M. Stahle, and B. H. Parker, “The effect
The Schottky-barrier height, , is a function of the applied of cathode bias (field effect) on the surface leakage current of CdZnTe
voltage and reflects the barrier lowering due to the voltage drop detectors,” Nucl. Instrum. Methods Phys. Res. A, vol. 510, pp. 300–308,
across the oxide layer. 2003.
The series resistance of the undepleted layer can be expressed [19] S. M. Sze, Physics of Semiconductors Devices. New York: Wiley,
1981.
as [20] E. J. Morton, M. A. Hossain, P. De Antonis, and A. M. D. Ede, “Investi-
gation of Au-CdZnTe contacts using photovoltaic measurements,” Nucl.
(A17) Instrum. Methods Phys. Res. A, vol. 458, pp. 558–562, 2001.

Authorized licensed use limited to: Indian Institute of TechnologyDharwad. Downloaded on March 15,2022 at 07:08:16 UTC from IEEE Xplore. Restrictions apply.
598 IEEE TRANSACTIONS ON NUCLEAR SCIENCE, VOL. 52, NO. 3, JUNE 2005

[21] A. A. Rouse, C. Szeles, J.-O. Ndap, S. A. Soldner, K. B. Parnham, D. [26] V. Gostilo, C. Budtz-Jorgensen, I. Kuvvetli, D. Gryaznov, I. Lisjutin,
J. Gaspar, M. H. Engelhard, A. S. Lea, S. V. Shutthanandan, T. S. The- and A. Loupinov, “The development of drift-strip detectors based on
vuthasan, and D. R. Baer, “Interfacial chemistry and the performance of CdZnTe,” IEEE Trans. Nucl. Sci., vol. 49, no. 5, pp. 2530–2534, Oct.
bromine-etched CdZnTe radiation detector devices,” IEEE Trans. Nucl. 2002.
Sci., vol. 49, no. 4, pp. 2005–2009, Aug. 2002. 2 2
[27] A. Shor, I. Mardor, and Y. Eisen, “Performance of 1 1 1 cm pixel
[22] M. A. Lampert and P. Mark, Current Injection in Solids. New York: CdZnTe gamma detectors,” IEEE Trans. Nucl. Sci., vol. 49, no. 4, pp.
Academic, 1970. 1935–1940, Aug. 2002.
[23] C. M. Hubert Chen, W. R. Cook, F. A. Harrison, J. Y. Y. Lin, P. H. Mao, [28] M. Amman, J. S. Lee, and P. N. Luke, “Electron trapping nonuniformity
and S. M. Schindler, “Characterization of the HEFT CdZnTe pixel de- in high-pressure-Bridgman-grown CdZnTe,” J. Appl. Phys., vol. 92, pp.
tectors,” in Proc. SPIE, vol. 5198, 2003, pp. 235–242. 3198–3206, 2002.
[24] C. Budtz-Jorgensen, I. Kuvvetli, N. J. Westergaard, P. Jonasson, V. [29] C. R. Crowell and S. M. Sze, “Current transport in metal–semiconductor
Reglero, and C. Eyles, “The x-ray imager on AXO,” Nucl. Instrum. barrier,” Solid-State Electron., vol. 9, p. 1035, 1966.
Methods Phys. Res. A, vol. 458, pp. 132–139, 2001. [30] C.-Y. Wu, “Interfacial layer-thermionic-diffusion theory for the
[25] A. E. Bolotnikov, W. R. Cook, S. E. Boggs, F. A. Harrison, and S. M. Schottky-barrier diode,” J. Appl. Phys., vol. 53, pp. 5947–5950, 1982.
Schindler, “Development of high spectral resolution CdZnTe pixel de- [31] H. C. Card and E. H. Rhoderick, “Studies of tunnel MOS diodes.
tectors for astronomical hard x-ray telescopes,” Nucl. Instrum. Methods I. Interface effects in silicon Schottky diodes,” J. Phys., vol. D4, p.
Phys. Res. A, vol. 458, p. 585, 2001. 1602, 1971.

Authorized licensed use limited to: Indian Institute of TechnologyDharwad. Downloaded on March 15,2022 at 07:08:16 UTC from IEEE Xplore. Restrictions apply.

You might also like