Hybrid pixel detectors: Characterization and optimization

(1)

Thesis for the degree of Doctor of Technology Sundsvall 2015

Hybrid Pixel Detectors

Characterization and Optimization

Erik Fröjdh

Supervisor:

Associate Professor Göran Thungström, Dr. Börje Norlin

Dr. Michael Campbell

Department of Electronics Design, in the Faculty of Science, Technology and Media Mid Sweden University, SE-851 70 Sundsvall, Sweden

ISSN 1652-893X

ISBN 978-91-88025-37-1

Mid Sweden University Doctoral Thesis 228

(2)

framläggs till offentlig för avläggande av doktorandexamen i elektroniktisdagen den 8:e september 2015, klockan 13.00 i sal O102, Mittuniversitetet Sundsvall.

Seminariet kommer att hållas på engelska.

Hybrid Pixel Detectors

Erik Fröjdh

©Erik Fröjdh, 2015

Electronics Design Division, in the Faculty of Science, Technology and Media Mid Sweden University, SE-851 70 Sundsvall Sweden

Telephone: +46 (0)10 142 80 00

Printed by Kopieringen Mittuniversitetet, Sundsvall, Sweden, 2015

(3)

dijjese

(4)

(5)

ACKNOWLEDGEMENTS

Having the opportunity to work on a PhD on hybrid pixel detectors at CERN, is something that I am truly grateful for. Not only for the inspiring location and the state of the art technology which I have been able to use, but also for the way in which it was realised. Always with enough guidance and sound scientific advise to keep me on track, but still having the freedom to pursue interesting questions that I encountered during my studies. For this I owe a lot to my supervisor Michael Campbell and my colleagues in the Medipix group. Thank you to: Jerome Alozy, Rafael Ballabriga, Xavier Llopart, Lukas Tlustos and Winnie Wong with whom much of this work have been done in collaboration with.

Also my gratitude towards Mid Sweden University for hosting me during the first two years of this work and supporting my move to CERN, with special thanks to my two supervisors in Sundsvall, Göran Thungström and Börje Norlin. David Krapohl deserves a mention as well, as much for our scientific collaboration, as for being a good friend.

The last three years of this work has been performed in the Marie Curie Initial Training Network, ARDENT, which was put together by Marco Silari in the RP group at CERN. Without him this work would not have been possible.

Thank you to Prof. Stanislav Pospisil for many interesting discussions on radiation detectors and for always showing great hospitality during my visits in Prague. To Erik Heijne for sharing his knowledge about pixel detectors and for organizing beam time at SPS.

This work is largely based on collaborative work and it would not be what it is without the contributions from many people, so thanks to: The team at the ANKA synchrotron in Karlsruhe, Thomas Koenig, Elias Hamann and Marcus Zuber for providing beam time and valuable feedback during the writing of our joint publications. Prof. Val O’Shea and Dima Maneuski at Glasgow University for sharing test beams and drinks. The staff in the detectors group at Diamond Light Source, Julien Marchal and Eva Gimenez-Navarro for a great collaboration and good test beam support. Stuart George for many discussions on pixel detectors as well as other things.

I believe that the success of the Medipix detectors is largely due to the strong collaboration formed around them and the open sharing of information among the different groups.

(6)

Also thank you to my dad Prof. Christer Fröjdh, for inspiring me to study physics and pursue a carrier in science.

Jerome Damet at CHUV-IRA in Lausanne for scientific collaboration and information on detectors and procedures for radiation protection. Claude Bailat also at IRA for good advice and sharp comments.

And of course all the other people including the CERN ski touring club, Gozo football team and all my friends that made my stay in Geneva to what it was.

This research has been partly supported by the Marie Curie Initial Training Network Fellow- ship of the European Community’s Seventh Framework Program under Grant Agreement PITN- GA-4 2011-289198-ARDENT

(7)

ABSTRACT

Hybrid pixel detectors offer an opportunity for improvements in several fields of radiation detection, from X-ray imaging to single particle tracking and personal dosimetry. The performance of the detector depends on several factors such as, radiation interaction in the sensor layer, charge transport and readout electronics. To optimize performance of the current generation of detectors as well as for the development of new detectors, knowledge of all these parts are vital.

This thesis covers several steps of the detection process in detail, and tries to give an overall picture of how they influence detector performance. For the sensor layer Cadmium Telluride and Cadmium Zinc Telluride was studied, looking at defects in the sensor and performance together with hybrid pixel detectors. Although the crystal quality have been improved in the last years, problems with defects affecting charge transport are still present. Hybrid pixel detectors need high quality sensors for optimal performance, but can also be used as a tool to characterize sensors for other applications using the pixellation to produce a detail map of the response.

Since many applications such as clinical CT imaging and quality assurance measurements on X-ray tubes lead to high incident fluxes on the detector, it is important to understand the high count rate performance of the detector system. Under these conditions pulse pileup in the front end can lead to lost counts and a distortion in the measured energy spectrum.

The high flux performance was studied using monochromatic radiation and a silicon sensor, in order to provide a clean input signal and focus on the pulse processing in the electronics.

Throughout the work simulations have been a valuable tool to complete measurements and improve the understanding of the detector response.

Because of the complexity of simulating a hybrid pixel detector careful design of the simulations is needed to provide sufficient accuracy in the process studied without giving unrealistic long run times.

(8)

(9)

SAMMANFATTNING

Hybrid pixel detektorer ger möjligheter till förbättringar inom många områden av strålnings detektering. Allt från röntgen till spårning av enskilda partiklar och dosimetri. Prestandan för detektorerna beror på många faktorer så som, strålnings interaktionen i sensorn, laddnings transport och utläsnings elektronik. För att kunna optimera prestandan för befintliga detektorer och för att utveckla av nya detektorer behövs kunskap om alla dessa steg. Denna avhandling tar upp vissa steg i detalj och försöker även ge en övergripande bild av hur de påverkar detektorprestandan.

Cadmium Tellurid och Cadmium Zink Tellurid sensorer har studerats i avseende på defekter och deras prestanda i kombination med hybrid pixel detektorer. Generellt har kvaliteten på sensor material förbättrats under de senaste åren men återstår fortfarande problem med defekter som påverkar laddningstransporten. Hög kvalitativt sensormaterial är viktigt för hybrid pixel detektorer men detektorerna kan även användas som ett instrument för att undersöka sensormaterialet, då pixelleringen ger en möjlighet att i detalj kartlägga responsen.

Eftersom många tillämpningar så som, datortomografi och kvalitetskon- troll av röntgenkällor leder till ett högt inkommande flöde av fotoner är det viktigt att förstå hur detektorn hanterar detta. Under sådana förhållanden kan överlappande pulser leda till missad detektering och en förvrängning av det uppmätta spektrumet. Studierna med högt foton flöde har varit fo- kuserade på elektroniken och därför har vi valt att genomföra dessa med monoenergetiska fotoner och att använda en kisel sensor även om inom många tillämpningar Cadmium Tellurid är ett bättre alternativ.

Genom hela arbetet har experiment kompletterats med simuleringar för att förbättra förståelsen för hur detektorerna fungerar. Vid simulering av hybrid pixel detectorer krävs tillräcklig noggranhet i de processes som ska studeras, men det är inte möjligt att fullt ut simulera alla steg i detekterings- processen eftersom det skulle leda till orimligt långa beräkningar.

(10)

LIST OF FIGURES

1.1 Schematic view of a hybrid pixel detector consisting of a semi conductor sensor layer and ASIC interconnected by a layer of bump bonds.

. . . 1 1.2 Pulse output from pre-amplifier with electronics noise and applied

threshold. Pixel logic working as implemented in Timepix. Note that TOA mode uses the closing of the shutter as reference. . . 2 2.1 Timepix chip with a 1 mm thick CdTe sensor. . . 8 2.2 Comparison of the single pixel mode and the charge summing mode

for a 2 mm thick CZT sensor with 110 µm pixel pitch for 25 keV photons from Sn XRF. The graphs have been rescaled to equal number of photons per unit area. . . 12 3.1 K-shell Auger and X-ray fluorescence yield as a function of atomic

numbers. Data from Krause [Kra79]. . . 18 3.2 Energy spectrum from 10 keV photons. Measured at the ANKA

synchrotron using Medipix3RX . . . 19 3.3 Energy spectrum from¹³⁷Cs measured with a 1 mm thick CdTe de-

tector bonded to Timepix. Measurement was performed TOT mode and clusters from one event summed. . . 21 3.4 Comparison of the components of the linear attenuation coefficients

for silicon and cadmium telluride . . . 23 3.5 Comparison of total linear attenuation coefficients for silicon and

cadmium telluride . . . 23 3.6 Fraction of photoelectric events for X-ray photons up to 150 keV . . 24 3.7 13 GeV/c Argon ion that fully traverses a 200 µm thick n-on-p silicon

sensor bonded to a Timepix3 chip. δ-rays clearly visible along the main track . . . 24 3.8 GEANT4 simulation of energy deposition per µm from 10 MeV alpha

particles in silicon. Sum of 1⋅ 10⁵particles . . . 25

(13)

List of Figures

3.9 Track from an Argon ion of 13 GeV/c traversing trough the sensor parallel with the sensor surface. Measured with a 500 µm thick silicon sensor coupled to Timepix3. Signal loss in the centre of the track is producing a rail road like pattern. . . 26 3.10 Track from an Argon ion of 13 GeV/c traversing trough the sensor

parallel with the sensor surface. Measured with a 200 µm thick n- on-p silicon sensor coupled to Timepix3. The signal in the centre saturates but it does not show the non monotonic behaviour seen in hole collection. . . 27 3.11 Track from an Argon ion of 13 GeV/c Track from an 13 GeV Argon

ion perpendicular to the sensor surface. . . 27 3.12 Clusters from¹⁰B - neutron interactions. Measured at INFN Legnaro

using a Timepix coated with a thin layer of¹⁰B. The incident beam was first thermalised using a thick block of polyethylene. . . 29 3.13 Sample clusters from the categories used in the neutron silicon exper-

iment, displayed ion a 70 x 70 pixel area. Colour scale represents the id of the cluster given by the clustering algorithm . . . 31 3.14 Neutron silicon interaction measured at LANSCE. . . 32 3.15 Statistical limit of the energy resolution in a silicon sensor for different

Fano factors . . . 33 3.16 Geant4 simulation of the deposited energy spectrum in an ideal planar

detector. 10⁶incident photos at 40 keV, 1 mm thick detector . . . . 33 4.1 Weighting field for a pixel with a width to height ratio 1:10 and centre

profiles for 1:3,1:5 and 1:10 as well as for a planar device . . . 37 4.2 Charge collection efficiency for a 2 mm planar device as function of

interaction depth. . . 39 5.1 Schematic view of the noise measurement. a) The detector is operated

in photon counting mode and pulses above the threshold is counted.

The threshold is scanned through the baseline and every point in b) represent one threshold setting. The result is then fitted with a Gaussian function to extract the noise. . . 42 5.2 Representation of an ideal threshold equalization using an 3 bit ad-

justment DAC. The symmetry of the Gaussian and the shift leads to a flat distribution after equalization. . . 43 5.3 Threshold dispersion at 8 keV in Medipix3RX before and after equal-

ization. . . 44

(14)

5.4 Correction of TOT gain dispersion in Timepix3 by single pixel calibration. Each line shows the measured²⁴¹Am spectrum for a single pixel, a) in TOT before calibration and b) in keV after calibration. . 45 5.5 Threshold dispersion for Medipix3RX measured with testpulses for

noise equalization and test pulse equalization at 8 keV. . . 46 5.6 Injected charge as function of pulse heigh. Measured using Medipix3RX

with a silicon sensor. . . 47 5.7 Peak position Cu Kα and a test pulse of equal charge. Mean distance

between the test pulse peak and Cu peak is 20e⁻or 0.07 keV. Note:

due to the minimum step of the pulse heigh the test pulse peak has been shifted 0.3 dacsteps . . . 48 5.8 Width of the test pulse peak (full chip) for noise, test pulse and test

pulse equalization with fine tuning. . . 49 5.9 Simulation of pulse pileup in the Dosepix front end. (Paper X, Simu-

lations by W. Wong) . . . 50 5.10 Dead time measurements for Dosepix with two different Ikrum set-

tings. (from paper VIII). Measured with monochromatic radiation of 40 keV with a threshold of 5.5 keV . . . 52 5.11 Simulations of the preamplifier output in Timepix for two different

input charges and ikrum settings. (Simulations by W. Wong) . . . . 54 5.12 Spectrum measurements with Dosepix in energy binning mode at the

ANKA synchrotron for 40 keV monochromatic radiation at various fluxes. . . 55 6.1 Simulation of the measured spectrum in a noiseless hybrid pixel

detector with a 300 µm thick silicon sensor with 55 x 55 µm²pixels subjected to 60 keV monochromatic photons. . . 58 6.2 Imaging simulation using Geant4Medipix set up for Medipix3RX

and using an imported CAD structure. Simulation by A. Schübel, Paper VII . . . 61 6.3 General structure of the Geant4Medipix framework . . . 61 7.1 By using a tilted geometry the depth of interaction is known from the

pixel coordinate. . . 64 7.2 Number of counts as a function of interaction depth for two beam

angles. . . 65 7.3 Spectra at different depths a the 1 mm thick CdTe sensor with 110 µm

pixel pitch. . . 66

(15)

List of Figures

7.4 Flat field maps of the two sensors studied. A. Blob defect, B. Single pixel defects, C. Line defects, D. Location of ring defect (only visible in hole collection) . . . 66 7.5 Raster scan of the blob defect. Arrows are drawn from the beam

position to the centre of gravity of the signal for each scan step. The colour scale refers to the total number of counts summed over all scan steps. . . 67 7.6 Scan pattern over the single pixel defect. The black dots represent one

scan step. . . 68 7.7 Cluster size and cluster TOT for a scan over a single pixel defect. Pixel

108 is the non responding pixel . . . 69 7.8 Measured gamma spectra for⁶⁰Co and²²Na using Timepix with a

1 mm thick CdTe sensor with 110 µm pixel pitch. . . 71 7.9 Single pixel spectrum from Sn fluorescence measured with the CZT

detector. Response is fitted with a modified error function.Note: This is the differential spectrum . . . 72 7.10 Energy resolution measured with indium fluorescence for 2 mm thick

sensors of CdTe and CZT. A 200 µm thick silicon sensor in included as a reference of frontend performance. . . 73 7.11 Dependency on bias voltage . . . 74 7.12 Flat field image from the 200 µm thick silicon sensor using Sn fluo-

rescence. Note that the colour scale is covering a 10x smaller range in respect to the expected standard deviation than for the high Z sensors. 75 7.13 Flatfield images taken with Sn fluorescence for the CdTe and CZT

sensor. The Poisson limit for the standard deviation is∼110 counts.

Both detectors show significantly higher deviations. . . 77

(16)

LIST OF TABLES

2.1 Overview of the Medipix ASICs . . . 8

2.2 Timepix Specifications . . . 9

2.3 Medipix3RX specifications . . . 10

2.4 Timepix3 specifications . . . 13

2.5 Dosepix specifications . . . 15

4.1 Charge transport properties of silicon, CdTe and CZT. . . 38

5.1 Front end noise ine⁻RMS for the Medipix detectors . . . 42

7.1 CdTe and CZT sensors tester with Timepix and Medipix3RX . . . 64

(17)

LIST OF PAPERS

This thesis is based on the following papers, herein referred to by their Roman numerals:

Paper I

Count rate linearity and spectral response of the Medipix3RX chip coupled to a 300µm silicon sensor under high flux conditions

E. Fröjdh, R. Ballabriga, M. Campbell, M.Fiederle, E. Hamann, T. Koenig, X.

Llopart, D. de Paiva Magalhaes and M. Zuber, Journal of Instrumentation, 2014 . . . 91 Paper II

Timepix3: first measurements and characterization of a hybrid-pixel detector working in event driven mode

E. Fröjdh, M. Campbell, M. De Gaspari, S. Kulis, X. Llopart, D. de Paiva T.

Poikela and L. Tlustos, Journal of Instrumentation, 2015 . . . 103 Paper III

X-ray absorption and charge transport in a pixellated CdTe detector with single photon processing readout

E. Fröjdh, B. Norlin, G. Thungström and C. Fröjdh, Journal of Instrumenta- tion, 2011 . . . 115 Paper IV

Imaging and spectroscopic performance studies of pixellated CdTe Timepix detector

D. Maneuski, V. Astromskas, E. Fröjdh, E.N. Gimenez, J. Marchal, V. O’Shea, G. Stewart, N. Tartoni, H Wilhelm, K. Wraight and R.M. Zain, Journal of Instrumentation, 2012 . . . 127 Paper V

Depth of interaction and bias voltage dependence of the spectral response in a pixellated CdTe detector operating in time-over-threshold mode subjected to monochromatic X-rays

(18)

E. Fröjdh, C. Fröjdh, D. Maneuski, E.N. Gimenez, J. Marchal, B. Norlin, V.

O’Shea, G. Stewart, H Wilhelm, R.M. Zain and G. Thungström, Journal of Instrumentation, 2012 . . . 139 Paper VI

Probing Defects in a Small Pixellated CdTe Sensor Using an Inclined Mono Energetic X-Ray Micro Beam

E. Fröjdh, C. Fröjdh, E.N. Gimenez, D. Krapohl, D. Maneuski, B. Norlin, V.

O’Shea, H Wilhelm, N. Tartoni, G. Thungström, R.M. Zain , Transactions on Nuclear Science, 2013 . . . 149 Paper VII

A Geant4 based framework for pixel detector simulation

A. Schubel, D. Krapohl, E. Fröjdh, C. Fröjdh and G. Thungström, Journal of Instrumentation, 2014 . . . 157 Paper VIII

Spectral response of the energy-binning Dosepix ASIC coupled to a 300 µm silicon sensor under high fluxes of synchrotron radiation

E. Fröjdh, F. Bisello, M. Campbell, J. Damet, T. König, E. Hamann, W.S.

Wong and M. Zuber, Submitted to NIMA, 2015 . . . 169 Paper IX

The Medipix3RX: a high resolution, zero dead-time pixel detector readout chip allowing spectroscopic imaging

R. Ballabriga, J. Alozy, G. Blaj, M. Campbell, M. Fiederle, E. Frojdh, E.H.M.

Heijne, X. Llopart, M. Pichotka, S. Procz, L. Tlustos and W. Wong, Journal of Instrumentation, 2013 . . . 179 Paper X

Measurement of an accelerator based mixed field with a Timepix detector S.P. George, C.T. Severino, E. Frojdh, F. Murtasa, and M. Silari, Journal of Instrumentation, 2015 . . . 197

(19)

Chapter 1

INTRODUCTION

1.1 Hybrid Pixel Detectors

Ionising radiation consists of a wide variety of radiation types ranging from ex- treme UV light, to relativistic heavy ions. A low energy limit of 10 eV is defined by the energy needed to ionise atoms in typical materials. [Kno10] Detection and measurement of ionising radiation is important in many different fields, from particle physics research to personal dosimetry and X-ray imaging. Ideally a radiation detector should be able to measure all types of radiation, over a wide range of particle energies and flux. The detector should also record important properties of the radiation, such as energy of the particles, time of arrival and spatial distribution. Many different designs of radiation detectors exist, for example ionisation chambers, high purity germanium detectors and thermoluminescent materials.

All detectors have their own strengths and weaknesses and it is important to choose a suitable detector for a specific measurement task.

The scope of this thesis is hybrid pixel detectors, which offer a unique combination of time, spatial and energy resolution as well as covering a wide range of flux and radiation types. The detector consists of a pixellated semiconductor sensor layer, bump bonded to a readout ASIC as shown in figure 1.1. Each pixel on the sensor has a corresponding pixel on the readout chip providing a dedicated pulse processing circuit. Using the hybrid approach allows for full area coverage and the possibility to optimise the sensor layer according to the radiation studied.

Following a radiation interaction in the sensor layer, charge generated by the ionisation drifts in the applied electrical field generating a small electrical

Read-out chip Sensor

Bump bonds

Figure 1.1: Schematic view of a hybrid pixel detector consisting of a semi conductor sensor layer and ASIC interconnected by a layer of bump bonds.

(20)

Time

Amplitude

Threshold Disc

1

Pulse Pixel Logic

PC TOT TOA Counter

4 5, 6....

Figure 1.2: Pulse output from pre-amplifier with electronics noise and applied threshold. Pixel logic working as implemented in Timepix. Note that TOA mode uses the closing of the shutter as reference.

pulse. This pulse is processed and measured by the electronics in the pixel. Figure 1.2 shows the pulse processing logic as implemented in the Timepix chip with photon counting (PC), time over threshold (TOT) and time of arrival (TOA) modes available. Having a large number of small pixels allows for precise position measurements as well as high flux capabilities due to parallel processing. By applying a threshold sufficiently high above the noise the detector records only events originating from radiation interactions in the sensor. Recording pulses and not integrating the total current also means that the detector is insensitive to dark currents and the low flux limit is determined by counting statistics and the natural background radiation.

The pixel pitch of state of the art multi purpose hybrid pixel detectors are in the range of 50-100 µm [Llo+07; Pan+07; Din+11; Bal+11; Bel+13; Poi+14]. For particle tracking and low energy X-rays silicon remains the most common sensor material while CdTe and CZT is the choice for medical imaging and gamma ray detection.

This thesis is based on measurements with the Medipix family of detectors and covers operation in photon counting, time over threshold and time of arrival mode.

1.2 The ARDENT Project

The author of this thesis has been funded through the ARDENT¹project which is a EU Marie Curie training network. It includes 15 Early Stage Researchers (ESR)

1http:://www.cern.ch/ardent

(21)

1.3. Summary of Publications and Personal Contributions

at 8 institutes spread across Europe and is coordinated by CERN and Dr. Marco Silari.

The aim of the project is to develop and test new technologies for radiation characterization and measurements of dosimetric quantities. Besides the scientific programme the ESRs are also offered extensive training opportunities. Hybrid pixel detectors are included as one of three main technologies explored in the project, the others being gas detectors and track detectors.

1.3 Summary of Publications and Personal Contributions

Paper I Count rate linearity and spectral response of the Medipix3RX chip cou- pled to a 300 µm silicon sensor under high flux conditions (Presented at the 15^th iWoRiD conference in Paris) includes measurements of energy resolution of the Medipix3RX detector with monochromatic photons in the range of 6 -15 keV.

It also presents measurements of count rate linearity and energy response under high flux conditions. The thesis author took part in the experiment at the TOPO-TOMO beamline at the ANKA synchrotron and performed the data analysis. Understanding the high flux response of hybrid pixel detectors is vital for applications in medical imaging.

Paper II Timepix3: first measurements and characterization of a hybrid-pixel de- tector working in event driven mode(Presented at the 16^thiWoRiD conference in Trieste) presents the first radiation measurements with the Timepix3 detector. The thesis author performed the measurements in collaboration with Xavier Llopart and analysed the TOT data. Per pixel analysis of the PC data was provided by Lukas Tlustos. The paper also describes and validates a calibration method using test pulses which is significantly faster than using radiation.

Paper III X-ray absorption and charge transport in a pixellated CdTe detector with single photon processing readout (Presented at the 12^thiWoRiD conference in Cam- bridge) characterizes a 1 mm thick CdTe sensor bump bonded to a Timepix chip.

It shows imaging performance, cluster size as a function of energy and bias voltage and detection of¹³⁷Cs where by using cluster summing the photo peak is recovered. The thesis author performed the calibration and measurements as well as the data analysis. The paper shows that thin sensors with small pixels can be used for gamma ray detection.

(22)

Paper IV Imaging and spectroscopic performance studies of pixellated CdTe Timepix detector includes a comparison between two CdTe sensors bonded to Timepix.

Both sensors were 1 mm thick with pixel pitch of 55µm and 110 µm respectively.

The thesis author participated in the experiments at the I15 beamline at Diamond Light Source and performed the measurement of the energy spectrum for the 110 µm detector. The paper provides information about how the pixel size affects the energy resolution of CdTe sensors.

Paper V Depth of interaction and bias voltage dependence of the spectral response in a pixellated CdTe detector operating in time-over-threshold mode subjected to monochromatic X-rays (Presented at the 13^thiWoRiD conference in Zurich) presents a study of the depth of interaction dependence and bias voltage dependence for energy spectrum measurements using 77 keV photons. The sensor was a 1 mm thick CdTe sensor bonded to Timepix and the experiment was performed at the I15 beamline at Diamond Light Source. The thesis author participated in the experiments and performed the data analysis. Validates that the response of a CdTe sensor with small pixels coupled to Timepix is not depth dependent. Also shows how a pencil beam can be used to characterize sensor material.

Paper VI Probing Defects in a Small Pixellated CdTe Sensor Using an Inclined Mono Energetic X-Ray Micro Beam (Presented at the 2012 IEEE/NSS/RTSD conference) investigate defects in a 1 mm thick CdTe sensor bonded to Timepix. By using a micro beam the response at different depths was investigated and the charge displacement due to the distorted electrical field visualized. The thesis author participated in the experiments at the I15 beamline at Diamond Light Source and performed the data analysis. Understanding how defects affect the detector response is important for development of sensor material and

Paper VII A Geant4 based framework for pixel detector simulation presents an extension to the Geant4 framework to simulate the complete signal chain in hybrid pixel detectors. The thesis author participated partially in the development of the code and provided knowledge about the Medipix3RX and Timepix detectors as well as data to validate the simulations.

Paper VIII Spectral response of the energy-binning Dosepix ASIC coupled to a 300 µm silicon sensor under high fluxes of synchrotron radiation characterizes the Dosepix chip regarding noise and energy resolution as well as high flux performance. The thesis author wrote the readout software used to control Dosepix,

(23)

1.3. Summary of Publications and Personal Contributions

led the experiments at the TOPO-TOMO beamline at the ANKA synchrotron and performed the data analysis. The paper also contains simulations of the pulse pileup in the frontend. Linking the measurement results with the frontend simulations validates the simulations for use in new designs.

Paper IX The Medipix3RX: a high resolution, zero dead-time pixel detector readout chip allowing spectroscopic imaging describes the Medipix3RX chip and present a first measurements using radiation. The author have been involved in verifying the function of the chip studying counter behaviour and stability under radiation.

The implementations of charge summing within chip with small pixels represents a major advancement for hybrid pixel detectors.

Paper X Measurement of an accelerator based mixed field with a Timepix detector discusses properties of the radiation field that can be measured from the mor- phology of clusters measured with Timepix. The thesis author assisted in the experimental setup and provided information about the Timepix detector. The author was also involved in the interpretation of the results. Linking radiation properties to track structure in the detector is important to extract the maximum amount of information.

(24)

(25)

Chapter 2

THE MEDIPIX FAMILY OF DETECTORS

2.1 Introduction

A spin off from high energy physics, the first Medipix collaboration was started to explore the possibility of photon counting X-ray imaging. It was composed of CERN, University of Freiburg, Glasgow University and INFN in Napoli and Pisa.

The first ASIC, Medipix1 [Cam+98], was a received in 1997. It was a prototype chip with 64 x 64 square pixels with a 170 µm pitch. The chip operated in photon counting mode with one analogue threshold and a 15 bit counter counter depth.

The initial collaboration was followed by the Medipix2 collaboration in 1999.

The aim of the new calibration was to produce a chip with smaller pixels and improved functionality. From the initial members the Medipix2 collaboration grew to contain 17 universities and institues. The Medipix2 chip[Llo+02] features a 256 x 256 pixel matrix with 55 x 55 µm²pixels. It has two analogue thresholds, a window discriminator and a 13 bit counter. A second chip, the Timepix [Llo+07]

was produced by the collaboration in 2006. This chip keeps the pixel size from the Medipix2 but in addition to counting photons above the threshold Timepix can also measure either the deposited energy using Time-over-Threshold, or the arrival time of the particle with the shutter as a reference. The Timepix chip will be described in detail later in this chapter and more information about the collaboration and applications of Medipix2 and Timepix can be found at cern.ch/medipix and in a review article by M. Campbell. [Cam11]

During the work with the Medipix1 and Medipix2 chip it became evident that the energy response of hybrid detectors with small pixels was limited by charge sharing. [CM01; Kor+07; Nor+06] A new collaboration was formed to design the Medipix3 chip [Bal+11] which implements analogue charge summing for 2 x 2 pixel clusters. After initial problems with the hit allocation to the charge summing nodes [Gim+11; Pen+11a] the fully functional Medipix3RX [Bal+13] was released in 2012. The latest chip from the Medipix3 collaboration is the Timepix3 [Poi+14;

Gas+14] chip which represents a major change, moving away from a frame based readout to data driven. Now immediately after a hit is processed by the pixel a data packet with information about arrival time and deposited energy is sent out. The Timepix3, has like the previous chips, a pixel matrix of 256 x 256 pixels with a pixel size of 55 x 55 µm²Even though it is the second Timepix chip the collaboration

(26)

agreed to name the chip Timepix3 to be consistent with the collaboration name.

Both the Medipix2 and the Medipix3 collaborations are currently active.

The Dosepix chip [Won+11] was designed in a research and development part- nership between CERN, Friedrich-Alexander University Erlangen-Nuremberg, and IBA Dosimetry. The chip has fewer (16 x 16) and larger ( 220 x 220 µm²) pixels than the Medipix detectors and is aimed at dosimetry and quality assurance of medical X-ray beams. The larger pixel size allows for more processing within the pixel and reduced charge sharing. Each pixel in the chip has 16 individually programmable digital energy bins, allowing for in-pixel spectroscopic measurements.

Table 2.1: Overview of the Medipix ASICs Chip Pixel Matrix Pixel pitch Year Technology

Medipix1 64 x 64 170 µm 1997 1 µm SACMOS

Medipix2 256 x 256 55 µm 2002 0.25 µm CMOS

Timepix 256 x 256 55 µm 2006 0.25 µm CMOS

Dosepix 16 x 16 220 µm 2011 130 nm CMOS

Medipix3RX 256 x 256 55 µm 2012 130 nm CMOS

Timepix3 256 x 256 55 µm 2014 130 nm CMOS

2.2 Timepix

Figure 2.1: Timepix chip with a 1 mm thick CdTe sensor.

The Timepix chip was first designed as a readout for TPCs [Llo+07] but has since then found many applications, ranging from time of flight mass spectrometry [Jun+11] and electron microscopy [Far09] to dosimetry in space [Tur+11b] besides being used for X-ray imaging. The pixel matrix consists of 256 x 256 square pixels at a 55 µm pitch. There is one globally applied analogue threshold and each pixel has a 14 bit counter. The specifi- cation of the chip are detailed in table 2.2.

(27)

2.2.1. Readout Systems

Table 2.2: Timepix Specifications

Pixel Matrix 256 x 256

Pixel Size 55 x 55 µm²

Counter Depth 14 bits Analogue Thresholds 1 Threshold Adjustment 4 bits

Threshold Dispersion ~35e⁻(adjusted) Operation Modes Photon Counting

Time over Threshold Time of Arrival TOT/TOA Clock 100 MHz (max) Electronics Noise ~100e⁻RMS

Timepix has three operation modes, photon counting (PC), time over threshold TOT) and time of arrival (TOA). The operation mode can be set individually for each pixel using the pixel configuration register. In PC mode the counter is incremented once for each time the preamplifier output crosses the threshold, while in the TOT mode the deposited energy is estimated by measuring the time that the pulse is above the threshold, incrementing the counter in the pixel once for each clock cycle that the discriminator is high. For the TOA mode the arrival time of the particle is measured relative to the closing of the shutter, as the pixel starts to count when the pulse crosses the threshold and keeps counting until the closing of the shutter. In both TOT and TOA mode each pixel can only accept one hit without distorting the measured data. For TOT mode the energy of a second hit would be added to the energy of the first hit and in TOA mode only the time of arrival of the first particle would be measured. A schematic view of the different modes in Timepix is found in figure 1.2 in the previous chapter.

2.2.1 Readout Systems

For the Timepix measurements in this thesis the Fitpix [Kra+11] readout system from IEAP/CTU was used. It is based on a FTDI FT2232H FPGA and uses USB 2.0 connection to the computer. The system has a maximum frame rate of ~90 frames per second and is fully integrated with the Pixelman [Tur+11a] software. Other readout systems for the Timepix chip also exists, like the USB key sized Fitpix Lite [VJ11] and the fast quad readout system, Relaxd [Vis+11] from Nikhef.

(28)

Table 2.3: Medipix3RX specifications General

Technology CMOS 130 nm

Measurement mode Photon counting

Readout type Frame based

Noise¹ SPM: ~100e⁻RMS

CSM: ~200e⁻RMS

Maximum count rate² 8.17⋅ 10⁷photons⋅ mm⁻¹⋅ s⁻¹ Fine pitch mode

Pixel matrix 256 x 256

Pixel pitch 55 µm

Thresholds SPM: 2

CSM: 1

Counters 2 x 12 bit

1 x 24 bit Spectroscopic mode

Pixel matrix 128 x 128

Pixel pitch 110 µm

Thresholds SPM: 8

CSM: 4

Counters 8 x 12 bit

4 x 24 bit

2.3 Medipix3RX

The Medipix3 chip is designed for spectral X-ray imaging, combining a small pixel size with an improved spectral response. The pixel matrix consists of 256 x 256 pixels with an intrinsic pixel pitch is 55 x 55 µm². There are two analogue thresholds per pixel and two 12 bit counters. Four different gain modes are implemented by selection of the feedback capacitance, in order to optimize the response in a specific energy range. There is also an option to bump bond every second pixel in x and y to combine resources from four pixels, which then gives an increased amount of counters and thresholds. The 55 µm pitch is referred to as fine pitch mode and the 110 µm pitch as spectroscopic mode.

1High gain mode and normal operating point

210 % loss in Fine pitch mode, Ikrum 100 10 keV photons. Details in paper I

(29)

2.3.1. Single Pixel Mode

2.3.1 Single Pixel Mode

In single pixel mode, each pixel works individually with two analogue thresholds and two 12 bit counters available. The counters can either be assigned to one threshold each, or combined to one 24 bit counter for extended dynamic range.

For the counters, 1 - and 6 bit operation modes are also implemented to offer higher readout speeds. Using the two counters with the same threshold there is also an option to operate the chip in continuous read/write mode (CRW), where one counter is read out while the other one is active. This gives a possibility to operate the chip with zero readout dead time, improving frame rate and giving an unlimited dynamic range. The CRW mode is also available in the charge summing mode.

In spectroscopic mode where one in four pixels are connected the pixel size is 110 x 110 µm²and eight thresholds per pixel are available. The switch between fine pitch and spectroscopic mode is programmable in the chip, but to have the correct charge collection area the a sensor with the appropriate pixel pitch has to be bump bonded.

2.3.2 Charge Summing Mode

When the chip is operated in charge summing mode, the charge is summed in 2 x 2 pixel clusters by analogue summing nodes and a network of arbitration circuits allocate the hit to the pixel with highest charge. By doing the summing using analogue circuits even pixels with charge below the threshold will contribute to the sum. The cost of the charge summing is a longer measured dead time per unit area and added noise (in quadrature) from the four pixels. In fine pitch mode the summing is done over 110 x 110 µm²and in spectroscopic mode over 220 x 220 µm².

The importance of the charge summing mode for detectors with small pixels can be highlighted by figure 2.2 which shows the energy response in single pixel mode and in charge summing mode for a 2 mm thick Cadmium Zinc Telluride sensor with 110 µm pixel pitch. In the single pixel mode the spectrum is totally dominated by charge sharing and no peak is visible while in the charge summing mode the photo peak is recovered. By recovering spectral information the charge summing mode also improves contrast over the single pixel mode and enhances material discrimination in spectral CT. [Koe+12]

(30)

10 15 20 25 30 35 40 Energy [keV]

0.0 0.2 0.4 0.6 0.8 1.0

Counts [1]

SPMSPM

Figure 2.2: Comparison of the single pixel mode and the charge summing mode for a 2 mm thick CZT sensor with 110µm pixel pitch for 25 keV photons from Sn XRF. The graphs have been rescaled to equal number of photons per unit area.

2.3.3 Readout systems

Given the broad range of applications several readout systems have been developed for Medipix3. These range from small and compact systems like the Fitpix [Kra+11] that was used for both the Timepix and Medipix3RX measurements in this thesis to faster systems like the Merlin [Pla+13], which can work at over 1000 fps and large area readouts for synchrotron applications like Excalibur [Mar+13], Lambda [Pen+11b] and Smartpix [Pon+15].

2.4 Timepix3

The Timepix3 chip [Poi+14; Gas+14] although sharing the same form factors as the previous designs, represents a major change, moving from frame based to data driven readout. Instead of reading out the full frame after the shutter is closed, in Timepix3 the data from one hit is sent off the chip immediately after being processed by the pixel. This dramatically reduces the readout dead time for low to medium hit rates, since only the TOT pulse time + 475 ns is needed for pulse

(31)

2.4.1. Readout Systems

processing and packet transfer from the pixel and during this time the rest of the chip is still active. It also means that the chip can be operated with continuous sensitivity under long measurements without digital pile up.

Timepix3 is designed in blocks of super pixels containing 2 x 4 normal pixels in order to share resources and save area. Each super pixel contains in addition to the logic for the packet transfer to the end of column a 640 MHz oscillator used for the fine time measurement. The oscillator is started when the pulse crosses the threshold and runs until the next 40 MHz clock cycle. In order to be able to send out the data packet immediately after the hit the TOA is not measured with reference to the shutter closing as in Timepix but instead a value is latched from a global TOA counter. At 40 MHz this counter rolls over every 410 µs but given that the rate is sufficiently low the readout system can allocate the packet to the right time segment and therefore not limit the total length of the measurement.

Table 2.4: Timepix3 specifications

Pixel Matrix 256x256

Pixel Pitch 55 µm

Measurement modes 10 bit TOT and 18 bit TOA, 18bit TOA only

10 bit PC and 14 bit integral TOT

Readout type Data driven

Frame based

(both modes with zero suppression)

Pixel dead time (data driven) > 475 ns (pulse processing + packet transfer) Maximum count rate 85.3 Mhits/s/chip

Minimum time resolution 1.56 ns

Power pulsing Yes

Minimum Threshold 500e⁻(~2 keV with silicon sensor)

2.4.1 Readout Systems

Currently there are two readout systems available for Timepix3, Spidr and Fitpix3.

The Spidr readout system developed at Nikhef is built around a Virtex7 FPGA and uses a 10GB Ethernet link for the communication. It is capable of reading out Timepix3 at full speed and offers an open source interface. Fitpix3 from

(32)

IEAP/CTU is instead aimed at portability offering a small form factor and only requiring a USB2 connection to the computer. For the Timepix3 measurements presented in this thesis the Spidr readout was used, except for the neutron silicon interactions measured at LANSCE¹where the Fitpix3 was used.

2.5 Dosepix

The Dosepix chip [Won+11] was designed in a collaboration between CERN, University of Erlangen and IBA Dosimetry. It is aimed at dosimetry in pulsed fields and quality assurance of X-ray tubes as a kVp meter. The pixel matrix consists of 16 x 16 pixels with a size of 220 x 220 µm². The larger pixel size was chosen to limit charge sharing and to accommodate a more complex digital pixel logic. Here the small pixels are also not needed for spatial resolution and the main reason for the pixellation is to provide many parallel channels. This way the flux per channel can be limited by having a small area per channel and the sensitivity can be increased by having many channels. A statutory requirement for dosimetric measurements is to be continuous sensitive, this is solved by reading out the chip one column at a time while the other columns are active.

2.5.1 Energy Binning Mode

The Dosepix chip can be operated in photon counting, and in integral TOT mode but for dosimetric measurements the energy binning mode is the most important.

In this mode the deposited energy is measured by time over threshold and the value is assigned to the corresponding digital energy bin. Each pixel has 16, 16 bit energy bins and with the bin energies individually programmable. Using energy binning allows for higher rates since the data does not need to be sent off the chip for each hit, and it lowers the requirement on the readout system since only the bin values need to be read out.

2.5.2 Readout systems

To control Dosepix and read out the data the Dosepix test board developed at IBA Dosimetry has been used. The test board is built around a EFM32 Cortex-M3 micro controller and supports ASCII commands from the PC. A new fully Python based readout software (dpxctrl) has been developed by the author. The software is designed to be flexible and allow a high level interface to the detector while

1Los Alamos Neutron Center, New Mexico, USA

(33)

2.5.2. Readout systems

still offering access to low level functions. It also offers the possibility to interface to other instruments using RS-232 or GPIB. Running in Python the user can access the full functionality of scientific tools as numpy, scipy and matplotlib for visualising and analysing data.

Table 2.5: Dosepix specifications

Pixel Matrix 16 x 16

Pixel Pitch 220 µm

Measurement modes Binned TOT mode with 16 x 16 bit energy bins PC Mode 8 bit

Integral TOT 24 bit

Readout type Frame based

Column based (with the rest of the chip active) Dead time 1.3 µs at 40 keV and Ikrum 15

0.61 µs at 40 keV and Ikrum 60 Maximum count rate 85.3 Mhits/s/chip

Minimum time resolution 1.56 ns

Power pulsing No, but sleep mode Minimum Threshold ∼5 keV with silicon sensor

(34)

(35)

Chapter 3

RADIATION INTERACTION IN THE SENSOR LAYER

A fundamental step in detecting ionising radiation, is the interaction of the radiation with the sensor layer. This might seem basic, but it is important to remember, as is sets the baseline for the performance of the full system. All steps after the interaction will seek to minimize the deterioration of the response and extract as much information as possible, but the performance can never be improved beyond this initial step.

Charged particles can be directly detected by the ionization of the sensor material while neutral particles like photons and neutrons are indirectly detected by their secondary products. In some cases where the interaction cross section is too low the addition of a converter material is needed.

3.1 Photon Detection

X-rays and γ-rays can interact with material in several ways but the important processes for radiation detection are: photoelectric absorption, Compton scattering and pair production. In each case the photon interacts with a single atom or electron in the sensor material fully absorbing it or scattering it through a large angle. This is in sharp contrast to the continuous slowing down of charged particles described later. Another important point to remember is that if the photon does not interact in the sensor layer, it will pass through without leaving any trace.

3.1.1 Photoelectric Absorption

During photoelectric absorption the incoming photon interacts with an atom in the sensor. The photon is absorbed and a photoelectron is ejected with the energy given by Eq 3.1:

E= hν − Eb (3.1)

whereE_b is the binding energy of the ejected electron and hν the energy of the incoming photon. The vacancy created can be filled either by capturing of a free electron or by rearranging electrons from the outer shells. During this

(36)

Figure 3.1: K-shell Auger and X-ray fluorescence yield as a function of atomic numbers. Data from Krause [Kra79].

process the excess energy is emitted as a characteristic X-ray photon, having an energy depending on the potential energy difference between the shells involved.

Additionally the energy can also be transferred to an electron in the outer shells ejecting it from the atom. This process is called Auger emission. Figure 3.1 shows the Auger and fluorescence yield in for an K shell interaction as a function of atomic number. Photoelectric absorption is also possible with M and L shells but with a lower probability. For an photon to interact with a certain shell it needs to have an energy higher than the binding energy of this shell, giving rise to the sharp absorption edges shown in figure 3.4

If the secondary products, including the photoelectron are stopped in the sensor, photoelectric absorption yields a single full energy peak thus making it an ideal interaction for spectroscopy and spectroscopic imaging. Figure 3.2 shows the recorded spectrum from 10 keV monochromatic photons using Medipix3RX at the TOPO-TOMO beam line at the ANKA synchrotron. The measurement was performed with a silicon sensor and the chip operating in CSM. The single peak is from photoelectric interactions and the width of the peak is a combination of

(37)

3.1.2. Compton Scattering

4 6 8 10 12 14

Energy [keV]

0.0 0.2 0.4 0.6 0.8 1.0

Counts [normalized, differential]

Data 10 keV Fit: FWHM 2.03keV

Figure 3.2: Energy spectrum from 10 keV photons. Measured at the ANKA synchrotron using Medipix3RX .

front-end noise, threshold and gain dispersion. Pixel detectors are more sensitive to fluorescence since the fluorescent photon only need to leave the pixel, and not the detector, to be detected as a separate event. Not only the yield (fig. 3.1) but also the energy and mean free path of the fluorescent photons increase with increased atomic number. Charge summing with inter pixel communication or energy and time measurements of each photon can be used to partially correct for this. Fluorescence is discussed in more detail in chapter 7 in the context of CdTe sensors.

3.1.2 Compton Scattering

In Compton scattering the incoming photon interacts with an electron in the sensor material and is scattered. In the scattering process a part of the photon’s energy is transferred to the electron which then can be detected. Depending on the angle the transferred energy can vary between close to zero and a large fraction of the photon energy. Even if the photon interacts and can be detected, the down side of Compton scattering is that the partial energy transfer makes it less suited for spectroscopic applications. However, as shown in equation 3.2, Compton

(38)

scattering carries information about the direction of the photon and gives, with a suitable detector the possibility to reconstruct the location of a radioactive source.

hν^′= hν

1+ hν

m0c²(1 − cosθ)

(3.2)

Figure 3.3 shows an energy spectrum from¹³⁷Cs measured with a 1 mm thick CdTe sensor with 110µm pixels bonded to a Timepix chip. Two photo peaks are visible, one from the primary 662 keV photons from Caesium and one peak at 32 keV from Barium fluorescence which is the decay product of Caesium. In addition to the photo peaks there is a major contribution from Compton scattering in the spectrum. The continuum is a result from having all different scattering angles in the sensor but two noticeable features are visible, the Compton edge and the Compton backscatter peak. The Compton edge is defined by the maximum energy that can be transferred from the incoming photon to the recoil electron. From eq 3.2 the difference in energy between the incoming photon and the recoil electron can be calculated as shown in eq 3.3

hν− Ee⁻∣_θ=π= hν

1+ 2hν/m0c² (3.3)

The backscatter peak is the result of Compton scattering in materials close to the detector. For scattering angles above 110°the scattered photon will have similar energies given the same primary energy. This will for monochromatic source result in a peak in the spectrum when these photons are absorbed in the detector.

Figure 3.3 also serves as a reminder of how important the radiation interaction in the sensor is and how different the measured spectrum can be from the incoming spectrum. In this case better readout electronics would only make the primary peak narrower not improve the overall shape of the measured spectrum. To improve the measured spectrum one would have to change the geometry and possibly implement a coincidence measurement to reject Compton events.

3.1.3 Pair Production

If an incoming photon has an energy higher than 1.02 MeV, which is twice the rest mass of the electron, it can interact in the sensor producing an electron positron pair. In this interaction the photon disappears and the excess energy is shared between the electron and position. As the position is slowed down in the sensor it

(39)

3.1.4. Linear Attenuation Coefficients

Ba kα

Compton backscattering

Compton edge

Photopeak

Figure 3.3: Energy spectrum from¹³⁷Cs measured with a 1 mm thick CdTe detector bonded to Timepix. Measurement was performed TOT mode and clusters from one event summed.

subsequently annihilates creating two 511 keV gamma rays. Just above the energy threshold this interaction is not very probably but for high gamma energies it becomes the most likely interaction. In this thesis since the energies used are relatively low no measurements are presented where pair production play a major role.

3.1.4 Linear Attenuation Coefficients

For a narrow beam of mono energetic photons with intensityI0passing through a layer of material with thicknessx and density ρ, the emerging intensity is given by:

I= I0e^−(µ/ρ)x (3.4)

whereµ is the linear attenuation coefficient. This is related to the interaction cross section (σ in barns/atom)by:

(40)

µ ρ = σ

uA (3.5)

whereu is the atomic mass unit and A the effective atomic mass of the material.

The interaction cross sectionσ is composed of the sum of the cross sections from the different photon interactions and it (or the linear attenuation coefficient) can be separated into the individual components. In figure 3.4 the linear attenuation coefficients of silicon and CdTe are plotted for the three different types interactions that are interesting for radiation detection and in figure 3.5 the sum of the coefficients are compared. At low energies photoelectric absorption dominates and as the energy increases Compton scattering and subsequently pair production takes over as the main interactions.

With a higer Z material such as CdTe the attenuation coefficients increase.

The probability for photoelectric interaction also extends further towards higher energies making high Z materials more suitable for spectroscopic imaging. This is not only important from a point of view of efficiency but also concerns the quality of the recorded spectrum since a greater fraction of the photons will deposit their full energy. In figure 3.6 the ratio between photoelectric absorption and total attenuation is plotted for silicon and CdTe over the energy range interesting for medical imaging.

3.2 Charged Particle Detection

The behaviour of charged particles is in strong contrast to photons. Unlike photons that can travel through a medium unaffected and then distinctly interact at a single point, charged particles continuously lose energy from Coulomb interactions.

When the particle moves through the medium the Coulomb force will either excite or ionise nearby atoms. The energy required is transferred from the charged particle.

If the particle passes close to an electron it can transfer enough energy that the electron can further ionise the sensor material. These electrons, called δ-rays are seen in figure 3.7 where an argon ion passes through a silicon sensor bonded to a Timepix3 chip

1NIST: XCOM Photon Cross Sections Database, http://www.nist.gov/pml/data/xcom/index.cfm

(41)

3.2. Charged Particle Detection

10³ 10² 10¹ 10⁰ 10¹ 10² 10³ 10⁴ 10⁵ Energy [MeV]

10¹⁰ 10⁸ 10⁶ 10⁴ 10² 10⁰ 10² 10⁴ 10⁶

Attenuation Coefficient [ mm−1 ]

Si, photoelectric CdTe, photoelectric Si, Compton CdTe, Compton Si, pair production CdTe, pair production

Figure 3.4: Comparison of hte components of the linear attenuation coefficients for silicon and cadmium telluride. Data from XCOM¹

10³ 10² 10¹ 10⁰ 10¹ 10² 10³ 10⁴ 10⁵ Energy [MeV]

10¹⁰ 10⁸ 10⁶ 10⁴ 10² 10⁰ 10² 10⁴ 10⁶

Attenuation Coefficient [ mm−1 ]

Si, total CdTe, total

Figure 3.5: Comparison of total linear attenuation coefficients for silicon and cadmium telluride.

(42)

0 20 40 60 80 100 120 140 Energy [keV]

0.0 0.2 0.4 0.6 0.8 1.0

Fraction of photo electric events

CdTeSi

Figure 3.6: Fraction of photoelectric events for X-ray photons up to 150 keV.

Figure 3.7: 13 GeV/c Argon ion that fully traverses a 200 µm thick n-on-p silicon sensor bonded to a Timepix3 chip. δ-rays clearly visible along the main track

(43)

3.2.1. Bragg Peak

3.2.1 Bragg Peak

Moving through the sensor the particle simultaneously interacts with many atoms.

The energy loss in a medium can be described with the Bethe formula:

−dE

dx =4πe⁴z²

m0ν² NZ[ln2m0ν²

I − ln (1 −ν² c²) −ν²

c²] (3.6)

wheree is the elementary charge, m0electron rest mass,ν and z the velocity and charge of the incoming particle. Z is the atomic number, N the number density and I the average ionization potential of the absorber material. Looking at eq. 3.6 we can see that for a non-relativistic particle the energy loss varies inversely with the particle energy. This gives rise to the characteristic energy deposition known as the Bragg peak. This is present for any charge particle stopping in the sensor material.

0 10 20 30 40 50 60 70 80 90

Depth [µm] 0

5000 10000 15000 20000 25000 30000 35000

Deposited energy [MeV/µm]

Figure 3.8: GEANT4 simulation of energy deposition per µm from 10 MeV alpha particles in silicon. Sum of 1⋅ 10⁵particles

(44)

Figure 3.9: Track from an Argon ion of 13 GeV/c traversing trough the sensor parallel with the sensor surface. Measured with a 500 µm thick silicon sensor coupled to Timepix3. Signal loss in the centre of the track is producing a rail road like pattern.

3.2.2 Effects from high charge deposition

Looking at equation 3.6 it is seen that the specific energy loss varies with the square of the particle charge. This leads to very high charge depositions for fully ionized heavy ions. In silicon a minimum ionizing particle deposits∼22 ke⁻in 300 µm [Par13], given simple scaling an Argon ion deposits more than 7 Me⁻in the same thickness and significantly more if it is stopped in the sensor. This charge will be spread by diffusion and repulsion but individual pixels still see charges in the range of 1 Me⁻. In addition the charge density will cause other effects, such as the plasma effect [WL74] delaying the charge collection time.

In the Timepix detector a distortion in the clusters from heavy ions have been observed [Gra+11; Hoa+12] and in recent experiments at the SPS beam line at CERN the same effect was also seen in the Timepix3 detector. In the central pixels that should record the highest charge the signal almost disappears as shown in figures 3.9 and 3.11. This is believed to be caused by a combination of sensor and ASIC effects but no comprehensive explanation has been given so far. In electron collection, with n-on-p sensors, we observe a saturation effect but not the strong attenuation of the signal in the centre of the track that is observed in hole collection. Figure 3.10 shows a track in a 200 µm thick n-on-p sensor from the same test beam as the p-on-n in the previous figure.

(45)

3.2.2. Effects from high charge deposition

Figure 3.10: Track from an Argon ion of 13 GeV/c traversing trough the sensor parallel with the sensor surface. Measured with a 200 µm thick n-on-p silicon sensor coupled to Timepix3. The signal in the centre saturates but it does not show the non monotonic behaviour seen in hole collection.

Figure 3.11: Track from an Argon ion of 13 GeV/c Track from an 13 GeV Argon ion perpendicular to the sensor surface.

(46)

3.3 Neutron Detection

Neutrons cannot be directly detected in silicon since they are neutral particles, but have to be detected through secondary particles, originating either from neutron silicon interactions or from interactions with a converter material. With a strong dependence on energy for the different interaction cross section, neutrons are classified by their energy as either fast or slow neutrons. Knoll [Kno10] uses 0.5 eV as the upper limit for slow neutrons and suggests that the lower detection limit for fast neutrons from recoil reactions is in the keV range. For both fast and slow neutrons the cross section in silicon is very low and despite giving a signal as shown in section 3.3.3 for any practical use the detector has to be covered with a suitable converter material.

3.3.1 Slow Neutrons

For the detection of slow neutrons several converter materials have been tested with the Medipix2 detector [Jak+06]. When choosing converter materials both the efficiency of the conversion and the detection of the secondary products is important. For example Gadolinium, despite having a very high slow neutron capture cross section, is not a suitable converter for neutron imaging with hybrid pixel detectors since the secondary particles are photons and electrons. A better option is either⁶L or¹⁰B where the secondary particles are short ranged heavy charged particles. Using charge sharing to precisely identify where the particle hit the sensor it is possible to achieve a spatial resolution in the order of one micron using 55 µm pixels. [Jak+09] While the short range of the secondary particles is beneficial for the spatial resolution it limits the efficiency since for a thick absorber the secondary particles will be absorbed in the converter. Simulations of a TiB₂converter on a planar silicon detector has shows an efficiency plateau for a thickness of about 2 µm. [Kra+12]. Figure 3.12 shows typical clusters from neutron capture reactions in a Timepix detector coated with¹⁰B.

3.3.2 Fast Neutrons

For higher energies the neutron capture cross section drops rapidly. Detection of fast neutrons with hybrid pixel detectors can instead be done using proton recoil in hydrogen rich materials [Uhe+08]. From the conservation of momentum the energy of the recoiling nucleus is given by:

(47)

3.3.2. Fast Neutrons

Columns

0 50 100 150 200 250

Rows

0 50 100 150 200 250

0 100 200 300 400 500 600 700 800 900 1000

Figure 3.12: Clusters from¹⁰B - neutron interactions. Measured at INFN Legnaro using a Timepix coated with a thin layer of¹⁰B. The incident beam was first thermalised using a thick block of polyethylene.

E_R = 4A

(1 + A)²(cos²θ)E_n (3.7)

whereA is the mass of the target nucleus divided by the neutron mass, E_nthe energy of the incoming neutron andθ the angle between the scattered nucleus and the incoming neutron. From equation 3.7 two important relations can be observed:

for a given converter material the energy of the recoiling nucleus depends on the angle of scattering and the neutron energy, and the maximum energy that can be transferred to the recoiling nucleus is given by the ratio of the nucleus mass to the neutron mass where only for elastic scattering with hydrogen the full energy of the neutron can be transferred. The detected energy depends then on the scattering angle and on how much energy the recoiling nucleus looses before entering the sensor.

(48)

3.3.3 Neutron - Silicon Interactions

To detect neutrons with a good efficiency the sensor needs to be covered a converter, as described in the previous section. However, it is important to understand the neutron silicon interactions in order to be able to separate them from the response from charged particles for accurate dosimetric measurements. Also a good knowledge about the interactions of neutrons in silicon could help to predict the risk for SEU events for electronics exposed to neutrons.

To characterize the neutron - silicon interactions experiments were carried out at the LANSCE²facility in cooperation with researchers from IEAP/CTU in Prague. A 300 µm thick silicon sensor bump bonded to the Timepix3 chip was placed in a neutron beam. Using the time of flight technique the energy of the incoming neutrons was known, and having the capability in Timepix3 to measure simultaneously energy and arrival time the response in the form of cluster shapes and deposited energy could be characterized. Figure 3.14 shows the number of interactions for different cluster types as a function of neutron energy. The cluster shapes give an idea of the particle generating the signal as discussed in paper X and sample clusters from the categories used in the experiment is found in figure 3.13 Dots in this case represents low energy transfer from the elastic scattering of the neutrons while blobs and heavy tracks represents nuclear interactions. In the response from the dots a clear edge can be observed showing a neutron energy threshold for ionizing interactions and also clear peaks from elastic scattering resonances.

3.4 Generation of charge carriers

In semiconductor hybrid pixel detectors the signal that is measured is the charge generated in the sensor from the incoming radiation. This process is fairly similar for the different types of radiation described since in the case of photons or neutrons the charge is generated by the secondary particles, often electrons. The energy required to generate on electron-hole pair varies between materials.

3.4.1 Fano factor

From an interaction in the sensor the energy is transferred not to a continuum of charge, but to discrete charge carriers in the form of electrons and holes. This

2Los Alamos Neutron Science Center, Los Alamos, New Mexico, USA

(49)

3.4.1. Fano factor

Dot Curly track Small Blob

Blob Heavy Track Light Track

Figure 3.13: Sample clusters from the categories used in the neutron silicon experiment, displayed ion a 70 x 70 pixel area. Colour scale represents the id of the cluster given by the clustering algorithm

process is subject to statistical variations and the fundamental limit on the sensors energy resolution is set by this variation. The fundamental limit set by Poisson statistics would be:

FWHM= 2.35

√N (3.8)

where N is the number of charge carriers. However, it has been shown that radiation detectors can achieve an energy resolution several times better than what is predicted by Poisson statistics. Using the Fano factor[Fan47] we can describe the variations in the amount of charge carries and how it deviates from Poisson statistics. The best possible energy resolution as a fraction of the peak energy then becomes:

FWHM= 2.35

√F

N (3.9)

Hybrid pixel detectors: Characterization and optimization