Skip to main content

Cooperation across Europe for Cd(Zn)Te based Security Instruments

Final Report - COCAE (Cooperation across Europe for Cd(Zn)Te based security instruments)

COCAE develops core technologies for nuclear radiation detection and its target outcome is a portable instrument able to answer where (within a range of some meters) and of what kind a radioactive source is.

The accurate answer to these questions is either time consuming and expensive or error-prone, until now.

COCAE utilises the Compton effect; a solution inspired by high energy astrophysics. Examples where the speedy and accurate acquisition of such information is valuable, apart from the case of a hypothetical terrorist attack, are the melting of a strong radioactive source in a furnace of a scrap metal recycling factory or the theft and the subsequent relinquishment of a source or the dispersion of a radioactive substance due to malfunction of a nuclear power plant.

The R&D within the collaboration was focused on the following core technologies:

1. The growth of high purity, detector grade Cd(Zn)Te crystals resulting in the development of the first 3 inch high quality ingots in Europe.
2. The fabrication of pixel detectors having structure of planar p-n and Schottky diode detectors with record energy resolution and Ohmic pixel detectors with wafer scale pixelisation.
3. The development of pixel electronics capable to record for each converted photon within the detecting layer its spatial coordinates, the time of conversion and the photon energy. A series of CMOS integrated circuits named Photon four-dimensional information (P4DI) were developed with last outcome the P4DI_v2F with 1250 pixels.
4. The hybridisation of the detectors and of the electronics and the design and construction of an instrument with purpose to explore the capabilities of the proposed method of radioactive source localisation and identification. Within the three and a half years of the project this target has been achieved partially, because the construction of the instrument has not been completed yet.

The design at all stages has been supported by a vast simulation effort at system level and by reconstruction algorithm development, both led by the application requirements. COCAE opens a pan European opportunity for:

1. implementation of a European based commercial activity for the supply of direct conversion Cd(Zn)Te crystals for x/gamma ray imaging applications with customers in Europe and worldwide;
and 2. for the utilisation of the crystals and associated electronics in commercialising a Cd(Zn)Te.

Compton camera system for identification and mapping of radioisotopes in various security environments such as:

- border security and inspections;
- security and inspections at recycling factories;
- nuclear waste management facilities and decommissioning of nuclear reactors;
- emergency response.

Project context and objectives:

The goal of COCAE was to develop core technologies which will enable the realisation of multifunctional radiation detection units capable to locate and identify radiation sources from emitted X and gamma rays in a broad spectrum of energies (up to 1.5 MeV). The main ideas which led the research effort were:

Compton imaging

Instruments like COCAE that exploit the Compton imaging technique deduce the energy of the incident gamma ray photons as well as their origin within a cone, by measuring the energy depositions and the positions of the Compton scattering interactions recorded in the detector. A typical design of such an instrument consists of two types of detectors, the scatter detector with relatively low atomic number, where the Compton scattering occurs, and the absorber with relatively high atomic number, in which the scattered photon is ideally totally absorbed. The COCAE instrument's CdTe detectors work both as a scatter, thanks to their arrangement into thin layers, and as an absorber, due to the large atomic numbers of Cd (S equal to 48) and Te (S equal to 52), resulting into a high photo-absorption efficiency.

Figure 1 illustrates the schematic of the Compton reconstruction principle. Considering a photon with energy incident on the detector's sensitive area that undergoes a Compton scattering, it will create a recoil electron of energy , quickly absorbed and measured by the detector, and a scattered photon of energy. The scattered photon ideally deposits its energy in the detector in a series of one or more interactions and is finally absorbed via a photoelectric interaction. The photon scatter angle is related to the measured energy depositions. Successive interactions of the emitted gamma rays create overlapping cones and the source location is the intersection of all measured cones (Figure 1(b)). In principle, three cones are sufficient to reconstruct the image of a point source. In practice, due to measurement errors and incomplete photon absorption, a large number of reconstructed cones are needed to derive the source location accurately.

CdTe – CdZnTe detectors

CdTe and CdZnTe offer a favourable combination of physical properties that makes them attractive as a room temperature radiation detector materials: The high atomic number (48,(30),52) and density (approximately 6 gr / cm3) lead to strong absorption of high energy photons and give high detection efficiency. Their relatively wide bandgap (1.5 - 2.2 eV) combined with electrical compensation techniques enables room temperature operation as very low carrier concentrations and leakage currents can be realised in these devices at room temperature. The relatively high mobility and lifetime of charge carriers, particularly those of electrons, allow the transport of the carriers through the whole depleted volume of mm and cm thick devices. However defects in the material cause carrier trapping and recombination which reduce the collected charge and make it dependent on the position within the crystal where the conversion of the photon to electrons and holes takes place. This deteriorates the energy resolution and the photo-peak efficiency. COCAE performed Research and development (R&D) in the quest for better (this means higher energy resolution, more stable in time), cheaper CdTe, CdZnTe detectors.

The improvement of the structural and charge transport properties of the material results to higher spectroscopic performance of the detectors. Consequently we tried to boost the Cd(Zn)Te crystal production quality and capacity by making various incremental improvements in two crystal growth methods, namely the Vertical Bridgman method and the Vertical Gradient Freeze method and by introducing 75 mm CdZnTe wafer production capacity with the Vertical Bridgman method and 50mm with the Vertical Gradient Freese method.

For cheaper detectors we tried circular wafer scale processing of Cd(Zn)Te. The semiconductor industry is equipped with production instruments that have been developed over the years for the multibillion Euro chip and memory fabs. These are of course based on the processing of 6, 8 and 12 Si wafers. Such production equipment can easily be modified or adopted as such for smaller sized wafers, for example 3, 4 and 5. Using these industry-wafer-scale production equipment results to high throughput and high yield, well controlled processes. Today even the leading manufacturers of CdTe (Japan) and CdSnTe (USA) are not processing circular wafers. Instead the wafer size and shape is asymmetric which requires specialised tools and non-standard processes, thus increasing the unit cost of each pixilated Cd(Zn)Te detector. The main reason to cut the ingot in asymmetric wafer shapes, is that the cutting direction impacts heavily on the performance through for example the impact it has on the crystal orientation. The crystals produced by ACRORAD have a (111) orientation which is optimal for the performance in x-ray imaging mode. Another reason is in order to select single crystal pieces out of multi grain bulk. Another direction is to make the transit time of the carriers through the device much shorter than their lifetime. This can be achieved by increasing the bias voltage of the detectors. If the leakage current becomes high the noise induced will cancel the improvement in energy resolution. Low leakage current results from both the high resistivity of the material and from the existence of a diode structure. The diode structures implemented are either Schottky barrier devices or shallow p-n junctions.

Indium is one of the favourable n-type dopants of p-type CdTe because it has a very shallow transition energy level and relatively low defect formation energy. However, the doping process of CdTe with this impurity is accompanied by the spontaneous formation of compensating opposite charged native defects or complexes of the dopant. In doping of CdTe achieved by equilibrium methods resulted hitherto in electron concentration not higher than 3 × 1018 cm3 because of the formation of compensating A-centers ((InCd VCd) pairs). We investigated the fabrication of p-n diodes using the method of laser induced solid-phase doping.

The ability to create metal-Cd(Zn)Te contacts (ohmic or blocking) with desired properties and which are mechanically sound and electrically stable is a key requirement in the fabrication of high energy resolution CdTe-based detectors.

Electronics and system development

Energy resolution, position resolution and time tagging capability are prerequisites for the operation of a Compton camera. This can be achieved only with VLSI multichannel signal processing electronics. The choices for such electronics are very limited world-wide and the existing chips are linear arrays with high power consumption. Electronics with 2D array of processing circuits and with low power consumption are photon counting, designed to serve X-ray imaging applications. We recognised the lack of appropriate electronics for the realisation of a Compton camera with a high number of readout channels.

Finally we undertook the endeavour to hybridise the pixel detectors with electronics by using bump and flip chip technologies and build a camera with 10 detecting layers and 100 000 individual readout channels. The detecting layers are placed 2 cm apart the one from the other and made of pixilated 2 mm thick Cd(Zn)Te crystals occupying an area of 4 cm × 4 cm. Each layer consists of a two dimensional array of pixels (100 × 100) of 400 micrometre pitch.

Project results:

We are developing a portable spectroscopic instrument which targets both the accurate identification and the localisation of radioactive sources and of radioactively contaminated spots based on high resolution energy measurements and on the use of the multiple Compton method.

Crystal growth

The first core technology addressed by COCAE was the growth of CdZnTe crystals. This was performed by the use of the vertical Bridgman VB and the vertical gradient freezing VGF methods.

The used source materials cadmium, sinc and tellurium were of highest available purity. Before the growth by VB or VGF the source materials were synthesised to improve the purity and to reduce the risk of explosion during the crystal growth experiments. For the growth high quality standard chemicals, high pure gases and semiconductor high quality quarts (HSQ 700 / 800) tubes were used.

Another very important part of the crystal growth experiments is the precise doping of the crystals. The crystals were doped with Indium to obtain high resistivity and high transport properties. The concentration of the dopants was rather small. There is only a small window for the right doping concentration of a few ppm to grow detector grade material. The balance between Indium donors and A centre's is considered the most crucial feature of the doping process. Only the right adjustment of the concentration gives detector grade material.

Vertical Bridgman method

The Vertical Bridgman Method, a method for growth from the melt, has been used. The raw materials cadmium, tellurium and zinc are synthesised in the same carbon coated quarts ampoule before the crystal growth experiment.

The temperature ramps for synthesis and homogenisation of CdZnTe have been presented which reach temperatures as high as 1250ºC for durations as long as 24 hours, nearly 150 degrees of Celsius above the melting point of CST. It is important, however, to consider that the softening point of quarts is already at 1 160 degrees of Celsius and such high temperatures for extended periods of time pose the risk of ampoule rupture. Thermo-mechanical methods for improving melt homogenisation have been proposed using oscillating temperature to cyclically melt and solidify the charge. CdZnTe homogenisation has been achieved using lower temperatures of 1160ºC in conjunction with a rocking furnace to break apart the clusters in the melt. Freiburger Materialforschungssentrum (FMF / ALU-FR) setup a crystal growth facility for the growth of 75mm diameter crystals (Error! Reference source not found.). Setup and parameters have been optimised with repeated growth experiments. Several CdZnTe crystals were grown by the vertical Bridgman Method (VBM) in the graphite coated silica crucibles. The CdZnTe crystals were doped with indium for a content of 10 percent sinc. Crystal ingots had 75 mm diameter and 1.6 Kg mass. A typical crystal (ingot) is shown in Figure 4(a). During crystal growth the temperature profile as a function of time is controlled. Two important features are monitored: The 'real' gradient during the growth and the exothermal reaction of the raw materials.

The ingots are cut to 1 and 2 mm thick wafers by a wire saw. The wire saw has proven to give the best results for cutting the wafers regarding resistivity and crystallinity. After cutting the wafers are polished, lapped and finally chemical-mechanically etched before material characterisation and detector processing.

Vertical Gradient Freeze method

Crystal growth Laboratory (CGL / UAM) constructed a vertical gradient Freeze furnace which gave the capability for growing 50mm diameter ingots. The design and the installation are shown in Figure 6.The furnace has five temperature zones and vertical and rotational support for the ampoule. The facility was used for the study of various ways to improve crystal and finally detector quality and produced several crystals

In the Bridgman and vertical gradient Freeze (VGF) techniques, it is common to use a sealed quarts ampoule for material synthesis and crystal growth of CdZnTe. However, to maintain the purity of the starting materials, it is important to prevent interaction between the quarts ampoule and the CdZnTe melt. One method commonly used is to deposit a layer of carbon on the quarts surface. The carbon layer is typically quite thin, on the order of microns. There is some doubt, however, whether or not this carbon layer effectively blocks quarts impurities from diffusing into the ingot. Other disadvantages include thermal stresses and interface curvature derived from using a quarts ampoule. A second approach is to use a high purity crucible, and to seal the crucible within the quarts vessel. One suitable crucible material which has been used with CdZnTe is pyrolytic boron nitride (pBN). One of the most commonly cited advantages of using pBN is the anisotropic thermal conductivity of the material. As a result of this, radial heat flow is suppressed. One disadvantage of pBN crucibles is associated with the crucible material interacting with the melt during the growth cycle. This observed interaction between the walls of the pBN crucible and that of the CdZnTe melt led to the development of a vacuum carbon coating system for applying a carbon coating to pBN crucibles.

Further experimental geometries and materials based on recent simulations of the thermodynamic characteristics of the CdZnTe growth process were tested and resulted to larger crystal grains and volumes, even from a region of the ingot, which is normally characterised as polycrystalline and low quality. The crucible was placed within a quarts ampoule whose inner surface was designed to mate to the outer surface of the pBN crucible. The ampoule was sealed under a vacuum greater than 1×10^6 Torr. The quarts ampoule with pBN crucible was placed upon a silicon carbine (SiC) pedestal with thermocouples arranged circumferentially. The implementation of a SiC ampoule support pedestal in conjunction with a conical pBN crucible geometry increases the axial heat flow, deflects radial heat flow, and improves temperature gradients at the crucible tip relative to other refractive materials studied. Furthermore, the geometry, which has been studied, leads to the formation of a convex Solid liquid interface (SLI) early on in the growth cycle.

The effects of surface treatment on the properties of CdZnTe have been studied. The material preparation operations such as cutting, lapping, and polishing create surface and sub-surface damage. To remove this damage layer, a chemical etching process is used. Chemical etching is needed prior to the metal deposition process in order to form a tellurium-rich surface, while removing surface contaminants and oxides.

Finally the effect of lateral surface morphology on detector performance was studied through the use of different lapping and polishing agents. It was found that polishing the edges of the planar detectors can help reduce leakage current by a factor of 200 % in most cases and improve the performance of planar detectors through the introduction of surface traps. These traps provide radiative recombination centres for surface travelling electrons.

Detector development

For the detecting layer to be used in the instrument we have studied the following alternatives:
1. Diode structures either p-n or Schottky prepared using commercially available CdTe crystals (by ACRORAD) and CdZnTe crystals prepared within the collaboration.
2. Ohmic type CdZnTe detectors using crystals developed within the collaboration. p-n diode detectors. Project partner ISP-NASU has developed and patented three different procedures of laser-induced doping of the surface region of Cd(Zn)Te.

The method used is the following: a rather thick (300-400 nm) In film is evaporated on the one side of a CdTe crystal. The In coated crystal surface is irradiated with a laser pulse (KrF excimer laser, 100 MJ / cm2, 20 ns pulse duration) in Ar environment of 0.3 MPa pressure and alternatively in water environment. The thickness of the deposited in film is such that it does not evaporate completely during laser irradiation and the film serves further as an electrode. The laser-heated plasma acts as a piston hitting the In film and generates a stress wave. The profile of the wave front becomes steeper and the stress wave is converted to a shock wave at a certain depth. Dopant (indium) atoms, implicated by laser-induced stress and shock waves, penetrate into CdTe. After laser processing a gold electrode is evaporated on the opposite side of the crystal. Laser-induced shock wave action can be considered as a stream of phonons which are scattered by the point and extended defects. The shock wave stimulates the dissociation of complexes, fast diffusion of impurity, desorption and segregation of interstitial atoms. In the case of CdTe crystals a large number of VCd and Cdi are formed as a result of the dissociation of (VCd-X) complexes and of the stimulated desorption of Cd atoms from the crystal lattice. In atoms are diffused at Cd vacancies and a large number of point donor defects InCd, Cdi and ClCd are frosen quickly without formation of the compensating acceptor complexes (VCd-X) like (VCd-InCd) and (VCd-ClTe). The doping occurs without heating the bulk of the crystal. In this way modification of the defect structure and of the properties of the bulk is avoided and thus a shallow and sharp p-n junction in the surface region of the CdTe crystal is obtained (resistivity of approximately 0-3 ??cm, electron concentration of approximately 1019 cm-3, mobility of approximately 140 cm2/V?s).

The production of CdTe p-n diodes using laser induced, solid phase doping has reached maturity with the work performed by the COCAE member ISP-NASU and the associated partner RIE. Hundreds of diodes were prepared and characterised. The yield of detectors with energy resolution better than 2 % at 662KeV was around 60% for samples prepared at the final stage of the work.

Schottky diode detectors

Another direction in the quest for higher energy resolution is the fabrication of Schottky diode detectors. We fabricated such kind of detectors with CdTe and CdSnTe crystals using Ni as the metal electrode for both contacts Schottky and Ohmic. Ni has high work function, while usually metals with low work function are used (In or Al) for the Schottky contact formation. Assuming the electron affinity of CdTe equal to 4.3 eV one finds its work function in the range of 5-5.7 eV (Eg equal to 1.46 eV). The preparation of ohmic contacts on p-type CdTe faces obstacles, because there is no available metal with work function value higher than the above. However, due to the ion bombardment of the surfaces the number of surface states is engineered and in the case of the 'nearly Ohmic' contacts small band bending is achieved, while with increased ion energy the density of surface states increases and larger band bending results. The crystal surfaces were etched chemically in a K2Cr2O7 + HNO3 + H2O solution for 20-30 s, they were cleaned, they were bombarded with Ar ions and finally Ni was deposited on them in a vacuum of 10-5 Torr. The procedures of fabrication of the Schottky and the Ohmic contacts differ only by the ion beam parameters: 'Ohmic' contacts were obtained with ion beam density 10-15 mA / cm2 and ion energy 400-500 eV, while Schottky contacts were formed with ion beam density 2-5 mA / cm2 and energy 700-800 eV.

In both cases, the etching time was about 10 min and the substrate was not intentionally heated (the temperature of the crystal was elevated no more than 2-3 oC due to efficient heat removal).

The Schottky diodes exhibit pronounced rectification properties starting from the lowest voltages, while the reverse current at 800 -1000 V is about two times lower than the best In / CdTe / Pt Scottky diode detectors referred in the literature and a steeper increase in the leakage current starts at higher voltages (at V > 1000 V instead of V > 200-300 V). The FWHM of such a detector is reported in Figure 10 (a), using a shaping time of 1.5 ?s, while the best spectrum recorded was obtained with 2?s shaping time and it is shown in Figure 10(b). The energy resolution figures achieved places the diodes prepared by CYFNU among the best in the world.

Detector surface pre-processing for nanocone formation

Project partner RTU has studied the mechanism of nanostructure formation on the surface of CdZnTe using laser radiation and the optical properties of these nanostructures. A new laser method was elaborated for cone like nanostructure formation on the surface of CdZnTe. The nanostructures begin forming at intensities I = 4 MW / cm2 on the irradiated surface. Different models have been proposed for the explanation of nanostructures self-assembly on a surface of a semiconductor by laser radiation. The thermogradient effect theory states that atoms with bigger effective diameter than atoms of the basic semiconductor material drift toward the maximum of temperature, while atoms with smaller effective diameter drift toward the minimum of temperature. The thermogradient effect theory explains the redistribution of Cd and Sn atoms at the irradiated surface of Cd1-xSnxTe at low intensities of laser radiation from 0.2 MW / ?m2 till 4.0 MW / ?m2. Two layers are formed near the irradiated surface of the semiconductor: the top layer consists of mostly CdTe crystal, while the underlying layer is a ZnTe crystal. A mismatch of lattices of CdTe and ZnTe crystals is equal up to 5.8 %.This plastic deformation of the top layer leads to the creation of nanostructures of the irradiated surface according to the modified Stransky-Krastanov mode. The energy of band gap of the Cd1-xSnxTe crystal increases along the axis of the nano-cone perpendicular to the sample surface. Thus, a graded band gap structure with optical window is formed in the nano-cone. A built-in quasi electric field, generated by graded band gap, is directed in the bulk of the semiconductor. As a result surface recombination velocity decreases and the mu-tau parameter value increased two times.

AJAT has developed a detector pixelisation process which is compatible with low temperature bump bonding technique. The process was developed initially based on single detector dummy samples. The wafer level pixelisation was considered a fundamental milestone, not only within the consortium but for Cd(Zn)Te industry worldwide. AJAT undertook the pixelisation process on the 3 inch CdZnTe wafers provided by ALU-FR. The experiment has been completed successfully, as can be seen in Figure 12.


The characterisation, being an integral part of the detector development program for COCAE, has advanced our understanding of how to prepare high energy resolution detectors. It helped to reach final decisions about the detectors to be incorporated in the instrument. Its results are highly recognisable by the room temperature semiconductor detectors development community.

An arsenal of tools was used for the characterisation of the electrical and detection properties of X and ?-ray detectors based on semi-insulating CdTe and Cd1?xZnxTe (x = 0.1-0.2).

A non-destructive method named contactless resistivity mapping (COREMA) (see Figure 13) together with infrared mapping (see Figure 14) were used for the production quality check of the crystal wafers.

The project partner CYFNU advanced the state of the art in crystal and detector characterisation by combining measurements and theoretical calculations. This work deepened our understanding of the properties of the material, which affect the detection capability of gamma rays by Schottky and p-n diodes.

It was proven that at low bias voltages the I-V characteristics of these diodes can be described by the the generation-recombination Sah-Noyce-Shockley theory. Furthermore it was proven that a decrease in carrier injection from the near-ohmic contact in the Schottky and p-n diode detectors is an important way to reduce the leakage current. Based on this, we succeeded to produce Schottky diode detectors with leakage current two times lower than the leakage current of the best available InCdTePt Schottky diodes at 800-100 V reverse bias voltage. It also helped to improve the behavior of the p-n diodes and allowed us to increase the operating voltage above 1200V.

The bandgap of crystals produced by the consortium members and by the Japanese company ACRORAD was determined. The method used was the measurement of the absorption coefficient of light by the crystals as a function of their wavelength for various temperatures. The band gap at 0 oK was extracted from the dependence of Eg with temperature.

The resistivity dependence on temperature was used to extract the thermal activation energy of resistivity and the Fermi level dependence on temperature. The Fermi level dependence on temperature was fitted in an expression resulting from the solution of the electro-neutrality equation.

The reliability of the method was confirmed by performing photoconductivity measurements for Schottky diodes prepared with pieces of the same crystals and by fitting them to an expression of their charge collection efficiency which took into acount the recombination losses in the diode space charge region. From that expression the value of Nd - Na was extracted as a parameter of the fit. The two independent methods led to remarkably same results.

One important conclusion from the above investigation was the following: A small variation of the concentration of compensating impurity not only changes the resistivity of the material but also can cause inversion of the type of its conductivity as the operating temperature of a device changes (see in Figure 15 that the Fermi level of CdZnTe-2 crosses the Fermi level of the intrinsic material). Apparently this can have a catastrophic impact on either a detector with ohmic contacts or with a Schottky contact when the operating temperature changes.

A second important conclusion relates the charge collection efficiency in a Schottky diode to the compensation degree of the semiconductor: There seems to be an optimum in the compensation degree for which the charge collection efficiency of the Schottky diode is maximum. Above and below this the spectral response of the device worsens. This was verified by gamma ray spectra measurements and it leads to the conclusion that a very high compensation degree which is considered beneficial for the increase of resistivity it destroys the spectral responce of Schottky diodes. In Figure 16(a) and (b) are presented spectra recorded with Schottky diode detectors prepared with crystals whose extracted parameters are recorded in table 2. One can notice that the crystal with highest concentration of uncompensated impurities gives the best response, while the almost totally compensated crystal is not able to record a spectrum. This result supports the theoretical calculation presented in (c).

This conclusion, as well as the technological complexities for the introduction of the diode processing steps in a mass scale production of pixel detectors within the time frame and budget limits of the project, led us to the decision to abandon the initial idea to produce pixel CdZnTe diode detectors and revert to Ohmic type ones.

VLSI electronics development

A large part of the COCAE effort was devoted to the development of the P4DI ASIC, a new generation of 2D imaging chips to be connected to a pixel sensor using the bump and flip chip technologies. The chip has been fabricated using a CMOS technology provided by the manufacturer UMC (from Taiwan) with 6 metal and one polysilicon layers and with minimum gate length 0.18 um.

The information provided by each pixel is a voltage level proportional to the charge delivered by the current pulse of the detector pixel and a voltage level proportional to the time interval between the moments when the current pulse starts and an external reference signal is asserted. These two voltage levels are digitised in pixel with 10 bits resolution and are sent outside the chip together with the digital pixel address. There are two important features which differentiate P4DI from the other existing ASICs, which have been used in the readout of Compton cameras. The one is the use of a Wilkinson type (analogue / digital) A / D conversion scheme for the in pixel digitisation of the dc levels which give the information of the peak amplitude and of the time the hit was recorded. The second is the readout mode: All the pixels are “open” in parallel to receive hits for a certain time interval adjusted externally. If a pixel receives a hit then it is disabled and cannot process any subsequent hit until the aforementioned time interval expires. After this recording phase the information is digitised, readout and then a new hit recording period can start. The benefits from these features are the following: The ASIC can process signals from radiation fields irrespective of their intensity without any pile-up. If the radiation field intensity is high the hit recording period of one frame can be shortened. Spatial patterns of bursts of hits can be identified. The ASIC is mixed signal but the digital activity is completely detached in time from the analog one. The analog signals, which are sensitive to noise pickup, are not transferred to the periphery of the ASIC. The power down of the analog part of the pixel after it has received a hit and until all the pixels of a given frame have been readout saves power. The disadvantages are a certain amount of dead time and the complexity of the power down circuitry.

It contains a charge amplifier which is directly connected to the detector through a bump bonding pad. A Leakage current compensation block is connected at both input and output of the charge amplifier. This circuit is responsible to sense and absorb the detector's leakage current so as to preserve the normal operation of the Charge amplifier. The charge amplifier is followed by the shaper amplifier, which shapes in time the produced signal and increases the signal to noise ratio, by filtering out noise components outside the bandwidth of interest. The readout scheme is designed in order to support detectors that either source or sink current or detectors with either holes or electrons as carriers. Thus, for detectors that source current the Charge amplifier is buffered and then connected to the Shaper, while for the case of detectors that sink current the carge amplifier's output is inverted and then driven to the Shaper. A comparator tracks the changes of the shaper’s output and produces a flag, when the pixel is hit. The hit flag activates both the peak detector and the time to voltage converter. The peak detector input is connected to the output of the shaper and stores the peak value of the signal. The time to voltage converter stores the time of the hit. The amplitude and time of hit information are then converted to a digital pulse by two The same information (amplitude and time of the hit) is also available in analog form. Using two buffers this analog information can be guided to dedicated output pads.

The layout view of one pixel and part of the pixel array are presented in Figure 18, while the P4DI_v2F ASIC and its precursors P4DI_v1 and P4DI_v2 are presented in Figure 19. The P4DI_v1 and P4DI_v2 ASICs were tested using dedicated printed circuit boards, FPGAs and software, which were developed using as interface to the PC a USB peripheral card. This card was developed and served as the core of the data acquisition of all the preliminary systems (see next section), as well as the archetype of the final data acquisition of the camera. The P4DI_v2 ASIC was flip chip bonded to 64 pixel detectors and tested using radioactive sources. The P4DI_v2F has been tested at wafer level using a dedicated probe card. Results from the test of the linear behaviour of P4DI_v2F for all gain settings are presented in Figure 20. In Figure 21(a) is plotted the resulting in-pixel A/D conversion digital code as a function of the charge injected to the pixel input. Spectra from an 241Am source recorded using a P4DI_v2 hybrid are presented in Figure 21(b).

System development

A prototype set up has been assembled as a stack of three PID350 pixilated gamma ray detectors, which were provided by partner AJAT Oy. The active area of the PID350 spans 4.5 × 4.5 cm2. each one detecting layer has 16 384 pixels. Each pixel is capable of storing the energy information of every detected event. Acquisition of data provides the user with histograms of event counts distributed in energy bins, but it is also possible to monitor and store single frames of detected events. A frame contains the energy amplitude of one detected event for each pixel. A data acquisition system has been developed for the readout of the stack of the PID350 detectors. The aim of the system was to increase the data transfer speed to the PC from 1 frame/sec, which was the rate of the PID350's standard interface, to 30 frames / sec. The continuous readout implemented in the PID350 pixel electronics can result to the inclusion of synchronous hits in different frames. Moreover the PID350 at its present status cannot assign a time stamp to a hit. These two reasons rendered difficult the identification of Compton events.

Nevertheless it was used for the validation of a radioactive source distance estimation algorithm. The algorithm estimates the detector-source distance s by using the photo-peak count information Ni from each PID350 detector layer i. Equation (1) is based on a model that takes into account the influence of solid angle and the absorption of the gamma rays by the detector layers. When the source is very near to the device, the number of converted photons is underestimated, because in one readout cycle the signal of only the last conversion in each pixel is read. Thus the algorithm 'sees' the source farther away that it is. When the source distance from the PIDs is considerably higher than the distance between the three sensing layers the sensitivity of the method is reduced.

The heart of the final COCAE instrument data acquisition system are the COCAE hybrids (Figure 25), which were manufactured by the partner AJAT using the P4DI_v2F ASIC and Ohmic type pixel detectors pixelated by Freiburger Materialforschungssentrum (FMF/ALU-FR) with CdZnTe crystals provided by FMF / ALU-FR, by CGL / UAM and by the Japanese company ACRORAD, while a number of detectors were bought directly from ACRORAD. More than 80 hybrids have been manufactured until now.

The mechanical drawing of the final system and the printed circuit boards of its subsystems are presented in the following figures. The COCAE instrument controller board is shown in Figure 26. It houses a FPGA which communicates using a point to point serial protocol with the 20 FPGAs, which gather data from the COCAE hybrids and upload to the P4DI_v2F ASICs their initial configuration values.

This card, being a USB2 peripheral, sends the hit data to the PC for reconstruction, radioactive source direction - distance estimation and identification. It can control through its powerful FPGA other subsystems. Subsystems that are foreseen are:

- An inertial sensor unit, which could provide data for the position of the instrument. These data are necessary for the implementation of developed triangulation algorithms, which estimate the radioactive source distance.
- A fan control unit, which could drive the ventilators with feedback from temperature sensors, which are embedded in the P4DI_v2F die, in order to create a stable thermal environment.

A High Voltage control unit for the bias of the detectors

The next board shown in Figure 27 controls one detecting plane. COCAE instrument has 10 such boards placed the one 2cm apart from the other. In its middle (where the 'Swiss cross' opening is located) the board shown in Figure 28 is plugged.

The hybridboard houses 8 COCAE hybrids, which constitute one detecting plane with 10000 pixels. The 'daughterboard' has on it two identical sub-systems, each one devoted to the control-readout of 4 hybrids. This is achieved with the aid of one FPGA for each sub-system. These FPGAs send/receive data to / from the motherboard using cables, which respect a low voltage differential signal standard. Furthermore they house ancillary circuits, such as digital to analog converters for the generation of bias voltages, ramp generation circuits which are necessary for the in-pixel A / D conversion and, finally, A / D converters which can be used to by-pass the in-pixel A / D conversion. The system mechanical enclosure was designed so as to fulfil requirements such as rigidity and robustness, ease of access for verification measurements, low mass for the minimisation of attenuation, cooling through convection, dark conditions inside the box.

Our current view of the final system specifications are collected in Table 3 and have resulted from our work in developing all the enabling technologies for it.

False alarm rate

Less than 1 per week of continuous operation at the test bed. False alarm is defined as a statistically significant (for example 95 % confidence level) excess of photons that is incorrectly identified by the device.

Simulations and data analysis algorithm development

The response of the instrument was modelled with the help of the Geant4 toolkit, that simulates the passage of particles through matter. All relevant physical processes (Compton scattering, photoelectric effect, pair production, electron / positron transportation into matter, and ionisation) were simulated. In particular, the Compton scattering was modelled accurately by taking into account the influence of the Doppler broadening effect (use of the package G4LECS). We used algorithms implemented in the open-source object-oriented software library MEGAlib which provides an interface to Geant4.

The algorithms used the spatial and energy information of each individual hit in order to form events (Compton scattering of photo-effect events). The reconstruction is split into two steps: a) Clustering, which means blobbing adjacent pixel hits into one larger hit, called a cluster. The energy of the hit of one pixel and of its eight closest neighbours are combined to form a cluster. The time tagging capability of the COCAE electronics permits to select hits with the same time tag. The energy of the cluster is the total energy of its pixels and its position is calculated via an energy-based center of gravity. b) Compton sequence reconstruction, which means that the clusters are sorted in the order in which the Compton interactions of the original particles occurred inside the detector. Multiple Compton scatterings can occur in the detector's sensitive materials before the scattered photon is ideally fully absorbed in the detector's volume. Since the distance between the detecting layers is of the order of 20 cm maximum, a time tagging resolution of some picoseconds would be needed to order in time the clusters belonging to the same event. This fact results to an ambiguity in the ordering of the Compton interactions sequence, which affects the correct determination of the Compton cone dramatically. If N clusters are recorded in the detector, there are N! possible sequences, of which only one is the correct one whereas the rest form a combinatorial background. We have studied different techniques that exploit the kinematical and geometrical information of Compton scattering events as well as statistical criteria in order to select the one with the best performance.

We have exploited two techniques: a) algorithms applied on events having only two interactions on the detector’s active volume, named Dual cluster sequence reconstruction (DCS) algorithms and b) an algorithm applied on events with more than two interactions, named Multiple cluster sequence reconstruction (MCS) algorithm. The simulation study proven that two DCS algorithms have the best performance: One that selects the combination with the highest product of the Klein-Nishina differential cross section multiplied by the probability for absorption via photoelectric effect and another which sorts the clusters assuming that the cluster having the highest energy deposition is the first Compton scattering. For events with more than two interactions the MCS algorithm assigns a Quality factor (QF), based on a generalised ?2 approach to each combination and selects the one with the minimum QF value.

The intrinsic detection efficiency of the detector is defined as the fraction of recorded events in the photo-peak over the total number of events crossing the first detecting layer. It has been studied as a function of the incident photon energy, by modelling a point source placed 80 cm from the detector’s centre (71 cm from the first detecting layer).

Amongst the benefits of the proposed instrument is its variable efficiency curve: It is possible to optimise the detection efficiency and the minimum detection limit imposed by the Compton plateau by activating only a part of its detecting layers. For example, if the identification of low energy gamma rays is of interest (like 241Am, 109Cd or the low energy part of the plutonium isotopes spectrum) the Compton plateau can be reduced by activating only one or two detecting layers of the instrument. Such a choice reduces the minimum detection limit and increases the sensitivity of the detector. Figure 30 shows the instrument's intrinsic detecting efficiency for different number of activated detecting layers as a function of the incident photons energy. It can be noticed that the intrinsic efficiency amounts approximately 10 % for a 600 keV radioactive source assuming that all the detecting layers are activated. As expected, the efficiency gets worse as the number of activated layers decreases and for higher energy incident photons.

The images of radioactive sources have been reconstructed by applying the List mode maximum likelihood expectation maximisation (LM-MLEM) imaging algorithm. According to this algorithm the image of a point source is generated by projecting each Compton event cone into an imaging projection sphere and then by performing successive iterations on the back-projected image in order to find the reconstructed image distribution with the highest likelihood of having produced the observed data.

Figure 31 depicts the reconstructed image for the case of an 800 keV point-like source located at 50 cm distance from the centre of the detector centre at azimuth angle and inclination angle (where corresponds to the detector symmetry axis).

The reconstructed image resolution of the COCAE instrument is defined as the FWHM of the reconstructed image distribution measured in steradian (sr) and it has been studied by considering two case conditions: point-like sources located a) on the detector’s symmetry axis (s) and b) off the detector's symmetry axis.

For the case of on-axis sources, the results of the simulations demonstrate an imaging resolution from approximately 2.5x10-3sr (for source-to-detector distances approximately 50 cm) down to ~0.5x10-3sr for point-like sources located at distances greater than 1m (Figure 32(a)). For the case of point-like sources located off the detector’s symmetry axis, it can be seen in Figure 32(b) that the reconstructed image resolution is less than ~4x10-3sr, for point-like source emitting photons with energies from 600keV to 2000keV. Figure 32(c) shows the reconstructed imaging resolution versus the inclination angle for 600, 1000 and 2 000 keV sources.

We have evaluated the reconstructed image resolution of the detector by varying the number of events used, in order to test the dependence of the performance of the detector on the number of recorded events. The minimum number of triggered events required for a Compton imaging is approximately 5?103.

Given the total efficiency of the detector (approximately 5-7?10-5 for point-like sources emitting photons with energies from 400 to 1 250keV located at s equal to 120 cm source-to-detector distance), the minimum detectable source activity has been estimated. Figure 33 illustrates results as a function of the data acquisition time. According to Figure 33, if the data acquisition time is 60 sec the system is able to detect radioactive sources down to 50 ?Ci activity. The ability of the COCAE detector to estimate the direction of radioactive sources has been studied.

Potential impact:

The dissemination and exploitation of the overall project results concentrated on two subjects, namely, the high energy resolution and efficiency image devices and the application of techniques to locate spatially and identify the kind of radiation sources in cargos as well as during emergency situations. During the project period the major activities that have been concluded are mainly related to the first mentioned subject (due to the delayed achievement of the final milestone) and have been widely disseminated.

Dissemination activities

A highly ambitious publications plan was implemented by the partners targeting high impact journals (e.g. Applied Physics Letters, IEEE Transactions on Nuclear Science, Nuclear Instruments and Methods in Physics Research, Materials Science and Engineering, Semiconductors Science and Technology, Journal of Applied Physics) and presentations to the IEEE NSS-MIC Symposium, to RTSD, to SPIE Meetings and other national and international conferences and symposia. With the active participation of the academic partners as well as the technology provider and technology user partners all of the objectives regarding the above quantified criteria have been met and exceeded.

All the results have been made available to the public through a well-maintained website, which includes a restricted area accessible to authorised users and a public section that describes the project and reports research highlights for the general audience. The project coordinator has registered the following web address in accordance to usual Seventh Framework Programme (FP7) practices: There a well-maintained website is hosted which includes a restricted area accessible to authorised users and a public section that describes the project and reports research highlights for the general audience.

During the project period numerous results have been presented in journal and conference papers which are listed below. It is worth noting that the list of publications in this section includes only the subset of papers officially acknowledging the COCAE project and / or funded through the project budget, while a larger number of publications reporting collaborative work among several of the COCAE consortium members and results that are outcome of project related activities have been published.

Exploitation activities

COCAE is a project whose results and deliverables open a pan European opportunity at two different levels: a) The possibility to explore, organise and implement a European based commercial activity for the supply of direct conversion Cd(Zn)Te crystals for X / gamma ray imaging applications with customers in Europe and worldwide and b) utilising such and associated electronics in a specific security application, namely commercialising a Cd(Zn)Te Compton camera system for identification and mapping of radioisotopes in various security environments. Without a doubt a European commercial activity on Cd(Zn)Te would serve a plurality of markets and applications (including dental, medical and NDT), but within the scope of the COCAE project the security application is one specific implementation of the technology.

Regarding the first target the market potential for Cd(Zn)Te according to an initial investigation and has been judged substantial and the technology has entered the market on three major sectors, namely dental, industrial and medical. The advantages of this new and novel technology are so significant that US, Japan, Canada, Israel have all established commercial supply of the material, components and sub-assemblies. Europe, although having participated in the R&D, still lucks a dedicated commercial facility for the production of Cd(Zn)Te. The COCAE consortium is promoting European excellence in the field of Cd(Zn)Te radiation imaging detectors and based on the experience of the partners the possibility of a pan-European centre for the production of high quality, pixelated Cd(Zn)Te radiation imaging detectors is real and would be a first class opportunity to leverage the know-how and the results produced by the project. This vision would require specific challenging follow up actions of this project, which have been laid out in (1). Regarding the second target for the exploitation of the project's Compton camera, a list of potential applications has been identified but a list of potential OEM (integrators) partners will be consolidated after the completion of the second major project milestone, although first contacts have already been made.

Specific plans and preliminary exploitation activities

In the frame of the COCAE project several important results have been achieved which show a huge potential for further exploitation. Therefore, the exploitation of the achieved results of the COCAE project is divided into several parts:

- crystal growth of CdTe and CdZnTe crystals;
- device processing using the p-i-n diode structure;
- semiconductor technology for processing pixel detectors with 75 mm wafers;
- read-out electronics for pixel detectors;
- exploitation of the complete system.

Obviously the above results have been accomplished due to the multidisciplinary nature of the project. To facilitate longer term collaboration in this direction the consortium partners have also collaborated on research proposals for funded activities and are actively investigating further possibilities to foster a longer term collaboration. Currently the consortium partners are investigating the possibilities of exploiting the technologies developed within the project through spin-off companies from the academia and collaborations with Small and medium-sized (SME) companies interested to invest, use and evolve the designs and processes developed by the project in their product lines as discussed in Deliverable D8.4.

Protection of Intellectual property rights (IPR)

The main rules for sharing IPR have been laid out in the consortium agreement, which had been finalised before the beginning of the project. Due to the discrete roles of consortium partners no conflicts or internal disputes have appeared. In order to avoid obstacles in the exploitation activities of the single partners after the end of the project, the participants will establish specific agreements regarding foreground knowledge sharing and special access to background know-how IPRs (if any), for ensuring a smooth transition to the post-CA phase.

List of websites: