Mid Sweden University

The charge transport and X-ray photon absorption in three-dimensional (3D) X-ray pixel detectors have been studied using numerical simulations. The charge transport has been modelled using the drift-diffusion simulator MEDICI, while photon absorption has been studied using MCNP. The response of the entire pixel detector system in terms of charge sharing, line spread function and modulation transfer function, has been simulated using a system level Monte Carlo simulation approach. A major part of the study is devoted to the effect of charge sharing on the energy resolution in 3D-pixel detectors. The 3D configuration was found to suppress charge sharing much better than conventional planar detectors.

Since the Medipix system is a photon counting system, it is possible to use it as a spectroscopic device, which is sensitive to the energy of the incoming photons. The system can, for example, be used to achieve the spectrum of an unknown source. Since it is an imaging system, it can be useful for energy dependent scattering experiments. The factors limiting the spectroscopic resolution are source and channel noise and detector non-uniformity. A well-known method for correction non-uniformity is to use mono energetic source and a threshold adjustment. However, since the threshold calibration is energy dependent, its use for sources with wider energy range is limited. We have shown an easy method to create threshold calibration masks for different thresholds that include both detector and channel stationary influences with analysis of histograms of series of images. The model for histograms was created assuming a standard dental X-ray source, which allows finding mask-creating parameters like threshold range (Vth and Vthadj). This procedure can be performed for each energy of interest. Based on these arrays, mask files that narrowed the threshold distribution close to the theoretical limit, were prepared. The limit of the spectroscopic resolution for the system was measured by analysing histograms for series of flat images under identical conditions.

An energy sensitive imaging system like Medipixl has proved to be promising in distinguishing different materials in an X-ray image of an object. We propose a general method utilising X-ray energy information for materialrecognition. For objects where the thickness of the materials is unknown, a convenient material parameter to identify is K = α¹/α², which is the ratio of the logarithms of the measured transmissions In(t¹)/In(t²). If a database of the parameter K for different materials and energies is created, this method can be used for material recognition independent of the thickness of the materials. Series of images of an object consisting of aluminium and silicon were taken with different energy thresholds. The X-ray absorption for silicon and aluminium is very similar for the range 40-60 keV and only differs for lower energies. The results show that it is possible to distinguish between aluminium and silicon on images achieved by Medipixl using a standard dental source. By decreasing the spatial resolution a better contrast between the materials was achieved. The resolution of contrasts shown by the histograms was close to the limit of the system due to the statistical noise of the signal.

The possibility to achieve better contrast in dental images by selecting specific X-ray energies is investigated. By applying both the low and the high photon counting energy threshold of the Medipix2 chip only a narrow interval of the spectrum of a 60 keV standard dental source used is selected. The relative contrast between the parts of the tooth with the highest respectively lowest absorption increases 18 % when the energy spread of the X-rays is narrowed from the full spectrum to a 4 keV window centered around 36 keV. The small tail of high energy photons from the full source spectrum lowers the achieved relative contrast significantly. This result can be used in development of X-ray sources or in consideration of energy weighting.

X-ray imaging with energy resolution can be performed using a detector matrix bonded to a photon counting CMOS readout circuit as the MEDIPIX2 chip. In previous experiments it has been shown that charge sharing between neighboring pixels plays an important role in the formation of the image and especially for the spectral information in the image. Charge sharing is caused both by the localization of the initial energy deposition and by diffusion during the transport of the charge to the readout electrode. In this work we have studied different factors that can effect the energy resolution in pixellated X-ray imaging detectors. Results are compared to experimental data.

A 1mm thick pixellated CdTe detector bonded to the MEDIPIX2 [1] readout chip has been characterized using a monoenergetic microbeam at the ESRF. This is an extension of the tests previously reported in [2]. The results show that a full energy peak can be obtained when a narrow beam is focused in the centre of the pixel. There is also evidence of significant charge diffusion and fluorescence. The results indicate that the charge sharing is the most important problem and will cause loss of the energy information in an imaging application. The second problem is the fluorescence which limits the number of counts in the full energy peak even for hits in the centre of the pixel.

Pixellated silicon detectors with a thickness of 300 µm and 700 µm bonded to the MEDIPIX2 readout chip have been characterised using a monoenergetic microbeam at the ESRF. The spectral response when a 10 x 10 µm2 wide 40 keV beam is centred on a single pixel is achieved. When the beam is scanned over the pixel, the charge sharing will increase when the beam approaches the border of the pixel. The experimental results have been verified by charge transport simulations and X-ray scattering simulations. Agreement between measurements and simulations can be achieved if a wider beam is assumed in the simulations. Widening of the absorption profile can to a large extent be explained by backscattering of lower energy photons by the tin/led bump-bounds below the detector. Widening of the detected beam is also an effect of angular alignment problems, especially on the 700 µm detector. Since the angel between the depth and a half pixel is only 2.2º, alignment of thick pixellated silicon detectors will be a problem to consider when designing X-ray imaging setups.

The ability to do X-ray imaging with energy resolution on a small pixel detector is severely limited by the charge sharing in the detector. Different schemes to overcome the problem have been proposed. Previous studies using synchrotron radiation have shown that, for a 300 µm thick Si detector, almost no charge sharing occurs for photons hitting the centre of each pixel. In this study we have used slits to focus the beam from a standard X-ray unit on the centre of a pixel on a MEDIPIX detector. The attenuation of the spectrum were measured for a number of samples of different materials with K-edge energy in the range of 30 to 50 keV. The measurements were done by scanning an energy window through the spectrum. Requirements for new X-ray imaging systems with true energy resolution are discussed based on these measurements.

In planar silicon detector structures the charge sharing between pixels is one limiting factor for colour X-ray imaging using integrated photon counting pixel detectors. 3D detector structures have been proposed as one solution to this problem. However, there are also readout system solutions to the problem i.e. introducing pixel to pixel communication and distributed charge summing in the readout electronics. In this work different charge summing schemes are evaluated using Monte Carlo simulation techniques. The increase in electronic noise introduced by the charge summing is one of the most severe problems. A proper selection of summing scheme is necessary to obtain an efficient system.