Development of high-Z sensors for pixel array detectors David Pennicard, DESY Heinz Graafsma, Sabine Sengelmann, Sergej Smoljanin, Helmut Hirsemann, Peter Goettlicher Vertex 2010, Loch Lomond, 6-11 June 2010

Development of high-Z sensors for pixel array detectors Applications of high-Z pixel arrays Overview of high-Z sensors CdTe / CZT GaAs Work on pixellated Ge sensors at DESY Summary

High-Z materials for X-ray absorption 7μm Fe, 2 mm H2O 400μm Fe, 4 cm H2O 4 mm Fe Notes – Si – Z = 14 GaAs = 31, 32, 33 CdTe = 48, 52

Synchrotron applications PETRA-III at DESY Beamline energies to 150keV (mostly 50keV) Materials science apps 3D X-ray scattering High-E scattering and tomography Structure at buried interfaces, grain mapping... One 150keV (material science) Focus the beam down to a small volume within a sample, measure X-ray scattering (e.g. from grain boundaries) Map out grain boundaries, region of different crystal structure. Both high quantum efficiency and high readout speed Precision measurement on bond length in amorphous materials and liquids Discuss hiZpad project!

Other applications Energy resolution! Medical & small animal imaging / CT Distinguish bone, tissue, contrast agents.. Astronomy Hard X-ray telescopes Gamma ray E.g. Compton camera... Johnson 2007 - Material differentiation by dual energy CT: initial experience Contrast agents – Iodine (33keV), barium (37.4keV) ASTRONOMY – Gamma rays for nuclear medicine and astronomy Already being used

Collaborations HiZPAD (Hi-Z sensors for Pixel Array Detectors) ESRF (coordinator), CNRS/D2AM, DESY, DLS, ELETTRA, PSI/SLS, SOLEIL CPPM, RAL, University of Freiburg FMF, University of Surrey, DECTRIS Medipix3 See Richard Plackett’s talk Inter-pixel communication allows thick high-Z sensors HiZPAD is EC funded – note list of European synchrotrons Don’t mention what it involves

Development of high-Z sensors for pixel array detectors Applications of high-Z pixel arrays Overview of high-Z sensors CdTe / CZT GaAs Work on pixellated Ge sensors at DESY Summary

General issues with high-Z sensors Fluorescence Degrades spatial and energy resolution above k-edge Bulk properties Leakage current, resistivity, trapping Material homogeneity and area Grain boundaries – want single crystal Pixellation Bump bonding 23.2keV fluorescence Note – photoconductor doesn’t need to be depleted - advantage Ejected Cd k- shell electron with 11.8keV 23.2keV deposited in other pixel

Cadmium Telluride Used for γ-ray spectroscopy Commercially-grown wafers: Single-crystal now 3” Defects affect uniformity Properties 1.44eV bandgap (room T) High resistivity Schottky or ohmic metal contacts Trapping & drift distances: Electrons - cm Holes - mm Use electron readout! Already Typically use Travelling Heater Method - see p-type resistivity from Cl compensation – 1e9 Ohm.cm µeΤe ~ 3*10-3 cm2/V µhΤh ~ 2*10-4 cm2/V Szeles 2003, CdZnTe and CdTe materials for X-ray and gamma ray radiation detector applications

CdTe Medipix2 Assemblies 1mm CdTe (Acrorad, 3”) Ohmic pixel contacts Te inclusions Hexa (2x3) 55 µm pixel pitch 28x43 mm2 active area,390,000 pixels Flat field & filter Produced by A. Fauler, A. Zwerger, M. Fiederle Freiburger Materialforschungszentrum FMF Albert-Ludwigs-Universität Freiburg QUAD (2x2) 110 µm pixel pitch 28x28 mm2 active area Flat field corrected

CdZnTe Typically Cd0.9Zn0.1Te Produced in large polycrystalline ingots Increased bandgap (1.57eV) – lower current Produced in large polycrystalline ingots Good single-crystal segments up to 20*20mm2 Small pixel arrays possible NuSTAR (Nuclear Spectroscopic Telescope Array) 20*20mm2, 600μm pixel size Production is by High Pressure Bridgman Note that better performance is variable – depends on how material is contacted Still potential, but not so suitable for small pixel tiled devices with no dead area Nustar – focusing, X-ray detecting satellite

Gallium Arsenide Better single-crystal production (6”) Response to monochromatic beam Better single-crystal production (6”) 1.43eV bandgap (room T operation) Problem – defects! Shallow defects prevent depletion Carrier lifetimes Epitaxial growth or compensation 90% CCE L. Tlustos (CERN), Georgy Shekov (JINR Dubna), Oleg P. Tolbanov (Tomsk State University) “Characterisation of a GaAs(Cr) Medipix2 hybrid pixel detector”, IWorid 2009 Attempting to produce „undoped“ GaAs is problematic, since it isn’t possible to get perfect stoichiometry with compound semiconductors. Shallow defects act as donors or acceptors, resulting in low resisitivity and hence difficult depletion. There is a common deep defect which occurs in GaAs produced from an arsenic-rich process called EL2 (arsenic atom occupying gallium site). This has a tendency to compensate for these shallow traps that could act as dopants. The result in semi-insulating GaAs, which has high enough resistivity to the used as a detector. (See S Watt’s thesis).   However, these EL2 traps tend to trap electrons, leading to a short electron lifetime (about 1ns given the conditions required for growing ingots). The solutions to this are to either: - use epitaxial growth at lower temp to reduce the no of defects - compensate DOPED GaAs with Cr to get suitable high-resistivity material GaAs (Cr) 300µm, ohmic contacts (Au)

Development of high-Z sensors for pixel array detectors Applications of high-Z pixel arrays Overview of high-Z sensors CdTe / CZT GaAs Work on pixellated Ge sensors at DESY Summary

Germanium pixels High-purity, high uniformity 95mm Ge wafers available Transport & depletion fine Narrow bandgap (0.66eV) means cooled operation needed Per pixel current must be within ROC limits (order of nA) Est. -50°C operation with Medipix3 (55µm) Need to consider thermal contraction, etc. “Engineering problems” Fine pixellation and bump-bonding must be developed

Pixel detector production at Canberra (Lingolsheim) Diodes produced by lithography (p-on-n) Thinned germanium wafer (0.5mm+) Li diffused ohmic back contact Boron implanted pixels Passivation, Al metallisation 2 runs planned: Medipix3 singles 55µm, 110µm and 165µm pixel, 500µm Second run 2*3 assemblies (42*28mm) Option of thicker Ge B (p+) Li(n+) M Lampert, M Zuvic, J Beau Information is available in Gutknecht 1990. Thickness of lithium layer? May affect efficiency at lower X-ray energies if large.

Bump bonding at Fraunhofer IZM (Berlin) Low temp bonding required Bonds must tolerate thermal contraction 3.5μm max displacement for ΔT=100K Indium bump bonding Bumps on ASIC and sensor Thermosonic compression at low T Possible reflow above 156°C Currently performing tests on Ge diodes Sputter coat TiW Photoresist mask Electroplate In Dissolve excess TiW Wafer dicing Flip chip assembly Optional reflow Note – infra-red detectors using GaAs quantum wells (smaller pitch than Medipix and same CTE mismatch) can be operated at LN2 temperature. So, should be OK. T Fritzsch, H Oppermann, O Ehrmann, R Jordan

Medipix3 module readout 2*6 chip module (28*85mm) Tilable Cooling through thermal vias Ceramic and heat spreader match Ge CTE Readout FPGA board 10 GBE for high-speed readout Improved infrastructure needed

Conclusions Demand for high-Z hybrid pixels Material science, biology / medicine, astronomy... Promising results from CdTe / CZT, GaAs Commercial CdTe / CZT wafers improving Improved GaAs compensation Ge pixels could provide high-uniformity sensors (albeit without room-temp operation)

Thanks for listening

What do hybrid pixels offer? Current generation (Pilatus, Medipix2, XPAD2/3) Noise rejection (photon counting) High speed Direct detection for small PSF Future detectors (Eiger, Medipix3, XPAD3+) Deadtime-free readout Inter-pixel communication (Medipix3) Correct for charge sharing Allows use of thick sensors Energy measurement Medipix3 provides 2 or 8 bins (55μm or 110μm) Medipix3 allows use of up to 1mm thickness without charge sharing losses

Choice of material and fluorescence effects Fluorescence harms performance immediately above k-shell ~26.7keV for CdTe ~11.1keV for Ge Motivation to use different materials 35keV photon in CdTe γ e- 23.2keV fluorescence γ e- CdTe – 23.17keV and 27.47keV respectively, compared to 26.7 and 31.8keV absorption edges, with mean lengths of 110um and 60um respectively. Note that the Te fluorescence can produce a second fluorescence from Cd k-edge (which is the most likely absorption) What about Ge? Edge is at 11.1keV. Photon is at EKα=9.88 keV , EKβ=10.99 keV. So, for k-alpha it is 50um. (Note that there are fewer electron shells, so “depth” of k-shell compared to other shells is smaller. e- Ejected Cd k- shell electron with 11.8keV 23.2keV deposited in other pixel

Choice of material and fluorescence effects Fluorescence harms performance immediately above k-shell ~26.7keV for CdTe ~11.1keV for Ge Motivation to use different materials 35keV photon in CdTe γ e- 23.2keV fluorescence γ e- CdTe – 23.17keV and 27.47keV respectively, compared to 26.7 and 31.8keV absorption edges, with mean lengths of 110um and 60um respectively. Note that the Te fluorescence can produce a second fluorescence from Cd k-edge (which is the most likely absorption) What about Ge? Edge is at 11.1keV. Photon is at EKα=9.88 keV , EKβ=10.99 keV. So, for k-alpha it is 50um. (Note that there are fewer electron shells, so “depth” of k-shell compared to other shells is smaller. e- Ejected Cd k- shell electron with 11.8keV 23.2keV deposited in other pixel

Cadmium Telluride Used for γ-ray spectroscopy Commercially-grown wafers: Single-crystal now 3”, 1mm-thick Defects affect uniformity Properties 1.44eV bandgap (room T) High resistivity Schottky or ohmic metal contacts Trapping & drift distances: Electrons - cm Holes - mm Use electron readout! Typically use Travelling Heater Method - see p-type resistivity from Cl compensation – 1e9 Ohm.cm M. Chmeissani et al. 2004, “First Experimental Tests With a CdTe Photon Counting Pixel Detector Hybridized With a Medipix2 Readout Chip” 23

Cadmium Telluride Typically use Schottky or ohmic contacts (Pt, Au, In) Temperatures above 200°C degrade transport properties Low temp sputtering / electroless deposition of contacts Low-temp bump bonding (Pb/Sn, In) CdTe relatively fragile Demonstrated with Medipix2, XPAD3 Medipix2 quad (FMF)

Gallium Arsenide Better single-crystal production (6”) 1.43eV bandgap (room T operation) Problem – defects! Shallow defects prevent depletion Carrier lifetimes Semi-insulating GaAs Compensation of shallow defects Operated as photoconductor / Schottky Epitaxial GaAs Growth with fewer shallow defects Operated as diode Attempting to produce „undoped“ GaAs is problematic, since it isn’t possible to get perfect stoichiometry with compound semiconductors. Shallow defects act as donors or acceptors, resulting in low resisitivity and hence difficult depletion. There is a common deep defect which occurs in GaAs produced from an arsenic-rich process called EL2 (arsenic atom occupying gallium site). This has a tendency to compensate for these shallow traps that could act as dopants. The result in semi-insulating GaAs, which has high enough resistivity to the used as a detector. (See S Watt’s thesis).   However, these EL2 traps tend to trap electrons, leading to a short electron lifetime (about 1ns given the conditions required for growing ingots). The solutions to this are to either: - use epitaxial growth at lower temp to reduce the no of defects - compensate DOPED GaAs with Cr to get suitable high-resistivity material GaAs (Cr) on Medipix2 JINR Dubna &Tomsk State University

Gallium Arsenide – Semi insulating As-rich growth produces deep defects (EL2) Compensate shallow traps But increase electron trapping (~100μm) Cr compensation promising Dope n-type during growth, then overcompensate p-type with Cr diffusion Metallised contacts Au for photoconductor (right) Pt-Ti-Au for Schottky Moderate temp tolerance, physically fragile Bonding at low temp Indium / low T solder 1ns lifetime, 1e7 cm/s max velocity implies 100um mean trapping length – not good JINR Dubna, Tomsk State University

Chromium-compensated GaAs Response to monochromatic beam Medipix2 300μm thick (1mm possible) Photoconductive sensor 90% CCE L. Tlustos (CERN), Georgy Shekov (JINR Dubna), Oleg P. Tolbanov (Tomsk State University) “Characterisation of a GaAs(Cr) Medipix2 hybrid pixel detector”, IWorid 2009 Anchovy head (flat field corrected)

Medipix3 256 * 256 pixels, 55µm pitch Photon counting Summing nodes 256 * 256 pixels, 55µm pitch 14.1 * 14.1 mm2 area Photon counting 2 counters / pixel (12bit) Continuous R/W or 2 energy bins Charge summing mode Optional 110µm pixels 8 energy bins 2000fps More with reduced counter depth Pixel Signal summing at nodes: Node with highest signal “wins”

Medipix3 256 * 256 pixels, 55µm pitch Photon counting Summing nodes 256 * 256 pixels, 55µm pitch 14.1 * 14.1 mm2 area Photon counting 2 counters / pixel (12bit) Continuous R/W or 2 energy bins Charge summing mode Optional 110µm pixels 8 energy bins 2000fps More with reduced counter depth Pixel Signal summing at nodes: Node with highest signal “wins”

Medipix3 circuitry 2 counters allow continuous read-write

Effects of charge sharing Loss of efficiency at pixel corners Typically, set threshold to E/2 with mono beam Loss of energy resolution Simulated pixel scan (500µm Ge) Simulated spectrum (500µm Ge) Counts lost in corner

Medipix3 charge summing mode Allows large sensor thickness while maintaining energy resolution No efficiency loss unless charge cloud > pixel size

Alternative methods of processing Ge Mechanical segmentation of contacts Frequently used for large sensors Limits on pitch Amorphous Ge contacts (e.g. LBNL, LLNL) Similar to Schottky Higher leakage current but allows double-sided strips