11th Workshop on Electronics for LHC and future Experiments, 12- 16 Sept. 2005 Heidelberg, Germany Pixel Detectors for Tracking and their Spin-off in Imaging Applications Norbert Wermes Bonn University LECC, 15.09.2005 Outline chip 1. Hybrid Pixels sensor „state of the art“ technology Æ tracking @ LHC Æ spin offs Æ SLHC Æ new ideas !? 2. (Semi) Monolithic Pixels the „future“ ÆILC driven Æspin-offs CMOS Active Pixels a-Si:H DEPFET Pixels SOI Pixels Bonn, Sept.5.-8. not covered: CCD development for ILC LECC-Heidelberg, 15.06.2004 Norbert Wermes, Bonn 1 State of the art … 10 yrs R&D Æ full detector Hybrid Pixel Detectors Bump & Flip-Chip Technology semiconductor pixel detector pixel contact ohmic contact detector substrates - Si Æ HEP - (diamond) - (GaAs) - CdTe Æ Imaging - CZT ATLAS (1.8 m2) my bias CMS (1 m2) ALICE (0.24 m2) (LHCb for RICH) flip-chip bump bonding pixel readout chip large (~m2) detectors are being built LECC-Heidelberg, 15.06.2004 NA60 (0.017 m2, running) Norbert Wermes, Bonn 2 Hybrid Pixels / LHC environment …. with particle fluence: 3 10 5000 1000 500 type inversion 100 50 10 5 102 ≈ 600 V 101 1014cm-2 "p-type" n-type 1 10-1 100 [M.Moll: Data: R. Wunstorf, PhD thesis 1992, Uni Hamburg] 0 10 1 10 2 10 Φeq [ 10 cm ] 12 -2 3 10 | Neff | [ 1011 cm-3 ] Udep [V] (d = 300μm) Change of Depletion Voltage Vdep (Neff) 10-1 2 1 • “Type inversion”: Neff changes from positive to negative (Space Charge Sign Inversion) before inversion p+ n n+ p+ p n+ 1 2 3 NIEL >1015 neq/cm2 dose > 500 kGy SLHC = LHC x 10 from M. Moll after inversion LECC-Heidelberg, 15.06.2004 Norbert Wermes, Bonn 3 Hybrid Pixels / Sensors pixel geometry ALICE Æ 50 µm x 400/425 µm (ATLAS, ALICE,NA60) Æ 100 µm x 150 µm (CMS) Æ thickness: 250 - 300 µm, ALICE: 200 µm CMS n+ pixels on n- oxygenated Si Æ after type inversion can operate partially depleted Æ bias grid to be able to test detectors before assembly Æ homogeneous charge collection 2 1 ATLAS charge collection eff. LECC-Heidelberg, 15.06.2004 bias grid Norbert Wermes, Bonn 4 Hybrid Pixels / FE-electronics Front End Chips • operation @ 40 MHz • zero suppression in every pixel • data buffering until trigger (µs later) 7.4 mm • low power (~ 28-90 µW / pixel) • e.g. ATLAS: amplitude via pulse width (ToT) • time walk for small signals • noise ~160e- (on module) • thres. dispersion ~600e (< 100e after tuning) • x-talk < 1% • thinned to 150 µm – 200 µm ATLAS 11 mm linear discharge ⇒ good columnar ToT architecture 1 mip threshold poor man’s analog R/O different injected charges ATLAS • several chip generations (1996-2003) radsoft: CMOS, BiCMOS (3-4 prototypes) radhard @ 1015neq /cm2 , SEU tolerant design • DMILL (yield too low), several iterations • 0.25 µm technology, several iterations • production finished (ATLAS: 250 wafers = 72000 IC) • chip yield on wafer ~ 50% - 85% LECC-Heidelberg, 15.06.2004 Norbert Wermes, Bonn 5 FE-chip wafer yields (0.25 µm CMOS) ATLAS (FE-I3) yields before thinning 82% 11 x 7.4 mm2 180µm thick LECC-Heidelberg, 15.06.2004 CMS (ROC) > 80% 7.9 x 9.8 mm2 200µm thick ALICE (SPD-RO) 51% 13.5 x 15.8 mm2 150µm thick Norbert Wermes, Bonn 6 ATLAS FE-I3: Pixel-Layout (Analogue Part) 400 μm 7-bit tune-DAC for local threshold adjustment Bonding-pad SEU-tolerant RAM cells for pixel configuration bits 50 μm Active voltage drop compensation Preamplifier 5-bit DAC – global threshold whole chip: 3.5 M transistors LECC-Heidelberg, 15.06.2004 RAM write/read control Auto-tune from I. Peric Norbert Wermes, Bonn 7 Important / in-time threshold & efficiency set threshold e.g. to 4000 ± 200 e(lowest possible ~1500 e-) In-time threshold (Δt<20ns): 5200 ± 200 e⇒ overdrive = 1200 ± 200 e- (achieved in irradiated assemblies) zero time walk line for large signals timewalk Î in-time efficiency ~99% wanted and achieved ! LECC-Heidelberg, 15.06.2004 Norbert Wermes, Bonn 8 Chip Architecture (animated) 40 MHz Gray coded clock transmitted to all cells Pixel circuit detects sensor signal (analog) and generates hit information (digital) Hit data with time stamps are temporary stored in end of column buffers outside pixel matrix The buffers monitor the age of each hit data and delete hits when no trigger coincidence occurs Hits having their time stamps (“expire date”) coincident with LV1 trigger are readout. Analogue circuits Digital readout circuits Registers used to store configuration bits Time information Trigger from I. Peric LECC-Heidelberg, 15.06.2004 Norbert Wermes, Bonn 9 Hybrid Pixels / hybridization Solder Indium 1. bumping 2. flip-chip (wafer scale) BARE module LECC-Heidelberg, 15.06.2004 Norbert Wermes, Bonn 10 Hybrid Pixels / hybridization bumping & flip chip of thinned bumped (!) chips (~ 180µm, 8“ wafers) 9 ATLAS (IZM) / ALICE (VTT) ATLASIndium (AMS) / CMS (PSI w. reflow) Solder (PbSn) 50 µm In reflow 50 µm Indium (lift-off) photo AMS, Rome photo AMS, Rome • „lift off“ + thermo compression • bumps „soft“ + „flat“ (~6 µm) - module handling more „touchy“ + can be done „in-house“ LECC-Heidelberg, 15.06.2004 50 µm photo IZM, Berlin • electroplating + reflow • automated wafer scale process @ vendor • bumps „strong“ and „taller“ (~25 µm) • very high yield if process steps well controlled Norbert Wermes, Bonn 11 Pixel Cross-Section 50 μm Sensor bumps (PbSn) 6 metal layers FE Chip ATLAS courtesy IZM LECC-Heidelberg, 15.06.2004 Norbert Wermes, Bonn 12 Hybrid Pixels / BARE module yield (ATLAS) 60% of produced @ IZM & AMS - ~ 2x20 modules/week - rework fraction : 10% - 15% - rework efficiency: solder ~100%, indium ~80% - module reject fraction: solder ~ 1%, indium ~14% 0.1% total need (3 layers): 1744 + spares total order @ bump vendors: ~2500 delivered (31.8.2005): 1500 fully assembled (today): ~1250 from J. Weingarten LECC-Heidelberg, 15.06.2004 Norbert Wermes, Bonn 13 Hybrid Pixels / pixel module MCC flex-hybrid sensor FE-chip pigtail out to opto interface bare d e l b m e s as LECC-Heidelberg, 15.06.2004 Norbert Wermes, Bonn 14 Hybrid Pixels / modules ALICE 1 sensor, 5 chips 12.8 x 7.0 cm2 ATLAS 1 sensor 16 chips 6.4 x 2.1 cm2 6.4 x 2.0 cm2 CMS (1 sensor, 16 chips) LECC-Heidelberg, 15.06.2004 Norbert Wermes, Bonn 15 Hybrid Pixels / module performance Production Test Sequence 1. digital functionality - injected vs. measured hits - injected vs. measured ToT - hits vs. column pair enable 2. analog functionality - hits vs. charge: threshold, noise meas., threshold tuning, crosstalk, timewalk - ToT vs. charge: ToT calibration & tuning 3. source test Am, 60 keV gamma - bump quality - dead channels - test ToT calibration 4. thermal cycling and tests at -10°C (ATLAS operating conditions) LECC-Heidelberg, 15.06.2004 Norbert Wermes, Bonn 16 Hybrid Pixels / threshold & noise tuned thresholds, dispersion < 100 e σ = 60e- noise ≈ 190 e mean = 190eRMS = 28efrom J. Weingarten LECC-Heidelberg, 15.06.2004 Norbert Wermes, Bonn 17 Hybrid Pixels / source scan (241Am) passive components on FLEX passive components on FLEX module burn-in tests 241Am γ - spectrum (no clustering) typically high pixel efficiency ~ few dead / 46080 (open/short) amplitude (via ToT) LECC-Heidelberg, 15.06.2004 from J. Weingarten Norbert Wermes, Bonn 18 Hybrid Pixels / module quality yield Ranking levels: b-layer, layer 1, layer 2 overall ranking total: 1225 ranking based on: - inefficient pixel - sensor quality - noise performance - threshold tuning - rebonding - BareModule rework failed: 120; 10% b-layer layer2: 256; 21% layer2 layer1 b-layer: 668; 54% layer2 failed layer1: 180; 15% layer1 b-layer from J. Weingarten LECC-Heidelberg, 15.06.2004 Norbert Wermes, Bonn 19 Hybrid Pixels / full pixel system tests BiStave in cool box “stand alone” and “module in system” difference distribution of • final power supplies • patch-panels threshold differences • final cables w original lengths PP0 with optoboard noise differences type3 cables Wiener LV supply σ = 12e σ = 50e ISEG HV supply from J. Weingarten LECC-Heidelberg, 15.06.2004 Norbert Wermes, Bonn 20 Main Issue for ATLAS and CMS: Irradiation target: 10 years LHC ≅ 1015 neq/cm2 ≅ 500 kGy • • • • • Si sensors: FE chips: glue: mechanics: cooling: depletion voltage and leakage currents threshold shifts & parasitic transistors becomes hard and brittle material performance degrades larger capacity needed to cool more power Î intensive irradiation and test beam program over years including dedicated high intensity beams with LHC like rates and timing structure LECC-Heidelberg, 15.06.2004 Norbert Wermes, Bonn 21 Irradiated Modules after 1 MGy (2xLHC) before irradiation noise thr <ENC> = 152e σthr = 40e LECC-Heidelberg, 15.06.2004 ATLAS 20yrs LHC lab measm’ts after 1 MGy noise thr <ENC> = 182e σthr = 127e Norbert Wermes, Bonn 22 Hybrid Pixels / Depletion Depth after 10 yrs Track position from beam telescope 600 kGy depleted non depleted 1/N*dN/dQ 18000e 0.07 0.06 Particle Track 200 V 400 V 600 V dN/dh The maximum track segment depth corresponds to an effective depletion depth (threshold dependent) 800 700 600 0.05 500 0.04 400 0.03 300 0.02 200 ATLAS 0.01 0 100 0 5 10 15 20 25 30 35 Cluster charge, all clusters 40 cluster charge LECC-Heidelberg, 15.06.2004 200 V 400 V 600 V 45 50 ke ATLAS 0 50 100 150 200 spurtiefe 250 300 350 track depth (µm) Norbert Wermes, Bonn depth (μm) 23 Hybrid Pixels / Trapping after 10 yrs LHC Also done using tilted tracks. For non-irradiated sensors,the collected charge is uniform along the depth. The charge yield yield as a function of the depth can be translated, via the drift velocity, in a carrier lifetime: Charge (arbitrary units) Charge Vs Depth 1 0.8 τe = 4.1 ± 0.3 ± 0.5 ns notirradiated 0.6 Irradiated 0.4 0.2 0 0 Bonn3, irradiated, 600 V (run1857) ATLAS 50 Bonn9, not irradiated, 150 V (run 1333) 100 150 mean CCE after 10 yrs LHC ~ 80% (with LHC type annealing scenario) due to 1. bias grid, 2. trapping 200 250 Track depth ( μ m) from A. Andreazza LECC-Heidelberg, 15.06.2004 Norbert Wermes, Bonn 24 Spatial resolution in irradiated assemblies before irradiation (100 incl.) after 60 Mrad (100 incl.) ATLAS from A. Andreazza LECC-Heidelberg, 15.06.2004 Norbert Wermes, Bonn 25 In-time track efficiency in irradiated assemblies efficiency before irradiation after 600 kGy efficiency no hit eff. 99.9% no hit 0.1% out of time 0% plateau 14ns 1 0.8 out of time 0.7% plateau masked 0.6 9.7 ns ~10-4 25ns 25ns 0.4 97.8% 1.5% 0.2 0 ATLAS 0 10 20 30 40 50 efficiency vs time, standard 60 70 ns large in-time plateau for efficency margin improved also by late hit duplication feature in FE-chip LECC-Heidelberg, 15.06.2004 Norbert Wermes, Bonn 26 Hybrid Pixels / Support Structures ALICE ATLAS ATLAS -minimal X0 “C-C” structures -cooling (pumped C3F8: bp -250) -T < -60 C to limit damage from irradiation -power dissipation: ~100W/stave (ATLAS) ~15kW/detector LECC-Heidelberg, 15.06.2004 Norbert Wermes, Bonn 27 Hybrid Pixels / Support Structure carbon-carbon space frame ATLAS ~1.8 m2, 80 Mpix LECC-Heidelberg, 15.06.2004 carbon-fibre structure CMS ~1m2, 50 Mpix carbon-fibre structure ALICE ~0.24 m2, 10 Mpix Norbert Wermes, Bonn 28 Main Issue for ALICE: minimal material In central HI collisions up to 8000 charged particles/ |η| are expected. ~80 hits/cm2 Primary charged particles in a central event from P. Riedler radiation levels only ~ 5 kGy, 6x1012 neq / cm2 Æ operation at room temperature possible ! LECC-Heidelberg, 15.06.2004 Norbert Wermes, Bonn 29 Main Issue for ALICE: minimal material ALICE very light weight Carbon Fibre support structure (200µm,0.1 X0) sensor 200µm IC 150µm cooling (C4F10) @ RT 0.3% X0 total X0 per layer ~ 0.9% (ATLAS, CMS > 2%) (PHYNOX tubes, wall 40µm) LECC-Heidelberg, 15.06.2004 Norbert Wermes, Bonn 30 NA60: LHC pixels put to test from M. Keil 8 (4 chip) + 8 (8 chip) planes Æ 12 tracking points in 2.5 T dipole ALICE1/LHCb pixel chip (@ 120 kGy) sensors operated through type inversion and through partial type inversion 107 ions/s → single hit rate up to 200 kHz for comparison: ATLAS B-Layer: 20 kHz LECC-Heidelberg, 15.06.2004 Norbert Wermes, Bonn 31 NA60: LHC pixels put to test from M. Keil addition of 4 planes Æ 16 tracking points in 2.5 T dipole 2004 pA running @ 2x109 p/burst 1 interaction / 25 ns Î upgrade to ATLAS production modules LECC-Heidelberg, 15.06.2004 Norbert Wermes, Bonn 32 Hybrid Pixels / NA60 results p-on-n type pixels become p-on-p after type inversion Lowering the bias voltage leaves a larger and larger undepleted region y [cm] After type inversion only fully depleted pixels should work 80 VV 100 150 50 60 40 30 V 222 4500 1800 2000 7000 4000 5000 1800 4000 1600 3500 6000 1600 3500 1400 4000 3000 1400 5000 3000 1200 1200 2500 2500 1000 4000 3000 1000 2000 800 2000 3000 800 2000 1500 600 1500 600 2000 1000 400 1000 400 1000 1000 500 200 500 200 1.5 1.5 1.5 111 0.5 0.5 0.5 000 -0.5 -0.5 -0.5 -1 -1 -1 -1.5 -1.5 -1.5 -2 -2 -2 -2 -1.5 -2 -2 -1.5 -1.5 -1 -0.5 -1 -1 -0.5 -0.5 000 0.5 0.5 0.5 111 1.5 1.5 1.5 222 [cm] xxx[cm] [cm] 000 from M. Keil LECC-Heidelberg, 15.06.2004 Norbert Wermes, Bonn 33 Hybrid Pixels / NA60 results Indium targets 23 MeV ΔMµµ position (mm) Identification of interaction vertex to 20 µm transverse, 200 µm along beam operation with ATLAS modules from M. Keil LECC-Heidelberg, 15.06.2004 Norbert Wermes, Bonn 34 Hybrid Pixels spin off into imaging applications LECC-Heidelberg, 15.06.2004 Norbert Wermes, Bonn 35 Hybrid Pixels for Imaging Applications radiography / crystallography film-foil system signal output digital imaging by counting of X-ray photons the principle • excellent linearity • large dynamic range • good image contrast • real time challenge build m2 & cheap detectors spin-off from HEP LECC-Heidelberg, 15.06.2004 dose / mGy under-exposed normal under-exposed normal over-exposed over-exposed Norbert Wermes, Bonn 36 Hybrid Pixels for Imaging Applications radiography / crystallography imaging by counting of X-ray photons the principle The challenges • high speed counting (107-9 cnts/s/mm2) >15 bit counter, > 1 MHz / pixel • no (little) dead time during exposure • operation at even lower noise / thr disp. ENC < 100e, σthr << 100e (tuned) • diff. energy window (double threshold) • high X-ray absorption (GaAs, Cd(Zn)Te) X-ray imaging • MEDIPIX Collaboration (CERN et al. ) • MPEC (Bonn Univ.) • Uppsala, Svedberg, Scanditronix Coll. really digital LECC-Heidelberg, 15.06.2004 Crystallography with synchrotron light XPAD (CPPM-Marseille @ ESRF) PILATUS (PSI @ SLS) Norbert Wermes, Bonn 37 Hybrid Pixels / counting pixels MEDIPIX (0.25µm CMOS) 256 x 256 pixels 55 x 55 μm2 max. count rate ~ 1 MHz/pixel 2 (3-bit tunable) discriminators 13 bit counter Si 14kV LECC-Heidelberg, 15.06.2004 MPEC (0.8µm CMOS) 2x2 chips, 1.3x1.3 cm2 2 x 2 chip modules (Si and CdTe) 32 x 32 pixels 200 x 200 μm2 max. count rate ~ 1 MHz/pixel 2 tunable thresholds, 18 bit cntr Si 20kV CdTe 60kV Norbert Wermes, Bonn 38 Hybrid Pixels / crystallography by counting PILATUS @ SLS , XPAD @ ESRF experimental setup Detector Plates Protein Sample monochromatic Xray Beam CPPM Marseille & ESRF protein crystallography diffraction expts (mat.sci) small angle scattering • Ephoton > 6 keV • spot size of diffraction maxima: 50 - 100 µm • Æ ~ 200 x 200 µm2 pixels, good PSF • single photon counting with large dyn. range • digestible hit rate per pixel = 1- 1.5 MHz • desirable: no dead area ! LECC-Heidelberg, 15.06.2004 Norbert Wermes, Bonn 39 Hybrid Pixels / crystallography by counting PILATUS @ SLS , XPAD @ ESRF experimental setup Detector Plates Protein Sample monochromatic Xray Beam PILATUS-1MCPPM @ SLS / PSI Marseille 20 x 24 cm2, 1 Mpix & ESRF Æ upgrade to 6Mpix protein crystallography diffraction expts (mat.sci) small angle scattering • Ephoton > 6 keV • spot size of diffraction maxima: 50 - 100 µm • Æ ~ 200 x 200 µm2 pixels, good PSF • single photon counting with large dyn. range • digestible hit rate per pixel = 1- 1.5 MHz • desirable: no dead area ! LECC-Heidelberg, 15.06.2004 from B. Henrich Norbert Wermes, Bonn 40 Hybrid Pixels / crystallography by counting PILATUS @ SLS , XPAD @ ESRF experimental setup CPPM Marseille &pixel ESRFB pixel A Detector Plates 50% Protein Sample critical threshold tuning for homogeneous response ! monochromatic Xray Beam • Ephoton > 6 keV • spot size of diffraction maxima: 50 - 100 µm • Æ ~ 200 x 200 µm2 pixels, good PSF • single photon counting with large dyn. range • digestible hit rate per pixel = 1- 1.5 MHz flat field image • desirable: no dead area ! LECC-Heidelberg, 15.06.2004 Norbert Wermes, Bonn 41 Thaumatin electron density Thaumatin crystal map Data Taking: Processing with XDS Data set: 120o Refinement with SHELXL Exp Time: 4s Integration: 1o Completeness: 90.3% Beam energy: 11.9 keV Beam intensity: 13.5% Rsym 8.4% D Sample-Det: 128 mm Resolution: 1.4 Å Resolution: 1.4 Å Analysis: Refinement: R-Factor 28% 3 data sets merged full geometrical correction Processed with XDS Robs: 8.9% (overall) Completeness: 90% (98% up to 1.6 Å) from B. Henrich Paul Scherrer Institut • 5232 Villigen PSI Beat Henrich Hybrid Pixels Hybrid Pixels challenges of a Super-LHC for pixels: Φ ~ 1016 neq/cm2 LECC-Heidelberg, 15.06.2004 Norbert Wermes, Bonn 43 RD50 Radiation Damage in Silicon Sensors A revie w in 5 s li des Two general types of radiation damage to the detector materials: • Bulk (Crystal) damage due to Non Ionizing Energy Loss (NIEL) - displacement damage, built up of crystal defects – I. Change of effective doping concentration (higher depletion voltage, under- depletion) II. Increase of leakage current (increase of shot noise, thermal runaway) 1016 neq/cm2 = III. Increase of charge carrier trapping (loss of charge) • Surface damage due to Ionizing Energy Loss (IEL) - accumulation of positive in the oxide (SiO2) and the Si/SiO2 interface – affects: interstrip capacitance (noise factor), breakdown behavior, … Impact on detector performance and Charge Collection Efficiency (depending on detector type and geometry and readout electronics!) 10 x LHC big effect M. Moll Pixel2005 Signal/noise ratio is the quantity to watch ⇒ Sensors can fail from radiation damage ! Michael Moll – PIXEL 2005, September 7, 2005 -44- RD50 Approaches to develop radiation harder tracking detectors Scientific strategies: I. Material engineering II. Device engineering III. Change of detector operational conditions CERN-RD39 “Cryogenic Tracking Detectors” Talks this Workshop H.Kagan, R.Stone D.Moraes S.Parker M.Swartz • Defect Engineering of Silicon • Understanding radiation damage • Macroscopic effects and Microscopic defects • Simulation of defect properties & kinetics • Irradiation with different particles & energies • Oxygen rich Silicon • DOFZ, Cz, MCZ, EPI • Oxygen dimer & hydrogen enriched Si • Pre-irradiated Si • Influence of processing technology • New Materials • Silicon Carbide (SiC), Gallium Nitride (GaN) • Diamond: CERN RD42 Collaboration • Amorphous silicon • Device Engineering (New Detector Designs) • p-type silicon detectors (n-in-p) • thin detectors • 3D and Semi 3D detectors • Stripixels • Cost effective detectors • Simulation of highly irradiated detectors • Monolithic devices Michael Moll – PIXEL 2005, September 7, 2005 -45- Hybrid Pixels / CVD diamond pixel detectors • ongoing effort RD42-ATLAS (Ohio-IZM-Bonn) • single chip modules (2880 pixels) and full ATLAS modules made • very nice results with 109Cd !! (22 keV, only 1600e) source 2250 2000 1750 1500 1250 1000 750 500 250 109Cd 0 0 4 6 8 10 12 Ladung/1000e position correlation Lab LECC-Heidelberg, 15.06.2004 2 Landau distr. Test Beam Norbert Wermes, Bonn 46 Hybrid Pixels / diamond pixel detectors what one can study very well with pixels … CVD diamond E T. Lari et al., NIM A 537 (2005) 581-593 diamond Si Look at correlation of the residuals of two events as a function of the track separation. LECC-Heidelberg, 15.06.2004 Norbert Wermes, Bonn 47 Hybrid Pixels / diamond pixel detectors Full 16-chip pixel module (using ATLAS FE-chips) Ohio IZM, Berlin Bonn RD42 – Ohio – IZM - Bonn LECC-Heidelberg, 15.06.2004 Norbert Wermes, Bonn 48 Hybrid Pixels / diamond pixel detectors noise map mean = 137e LECC-Heidelberg, 15.06.2004 threshold map tuned sigma = 25e Norbert Wermes, Bonn 49 Testbeams: CERN (4hrs !) and DESY (4 GeV) CERN CERN residuals DESY 1 chip module 16 chip module 7µm resolution multiple scattering dominated ! homogeneity tests under study LECC-Heidelberg, 15.06.2004 -30µm +30µm -150µm Norbert Wermes, Bonn 37µm +150µm 50 Hybrid Pixels / Edge Active 3D Detectors Electrodes are processed inside the detector bulk instead of being implanted on the wafer's surface. Holes are plasma-etched and then filled with doped polysilicon. • charges drift parallel to surface over short range Æ fast and radhard • the edge itself is an electrode Æ large active area (close to beam !) f • lower fields for full depletion Æ less breakdown detector photo speed: planar 3D 4. 4. from S. Parker 4. LECC-Heidelberg, 15.06.2004 Norbert Wermes, Bonn 51 Hybrid Pixels / Edge Active 3D Detectors pixel detector assemblies are still to be made ! first assemblies with bump bonded ATLAS FE-I3 did not work ! effort is repeated very soon ! LECC-Heidelberg, 15.06.2004 Norbert Wermes, Bonn 52 (Semi)-Monolithic Pixels a „dream“ of detector physicist 1. (high Z) semiconductor sensor with 2. fully integrated ampl. circuitry and R/O logic using 3. commercially available CMOS technologies developments largely driven by requirements for the ILC Total > 500 MPixel with ~ 25x25 µm cells hit rate = 80 hits / mm2 / bunch train (due to beamstrahlung e+e- pairs) occupancy ~20% Î need tricks eg. permanent R/O to both sides LECC-Heidelberg, 15.06.2004 Norbert Wermes, Bonn 53 CMOS Active Pixels LECC-Heidelberg, 15.06.2004 Norbert Wermes, Bonn 54 Monolithic Pixels / CMOS Active Pixels charge coll. in several µm thin epi-layer by thermal diffusion to n-well/epi junction p-wells and substrate highly doped Æ charges kept between reflection boundaries signals processed by standard CMOS circuitry integrated on sensor only nMOS in active area (due to n-well/epi collection diode) Q-collection time ~100 ns (due to diffusion) incomplete Q-collection and small signals (< 1000 e) => challenge for IC design small pixel sizes (< 20x20 µm2): a must and a virtue ! 1 Pixel - Cluster Signal Distribution MIMOSA I 4-diode pixel Nent: 15721 Big Peak Mean: 55.53 SD: 17.53 R2: 3.74 Second Peak Mean: 109.09 5.9 keV SD: 3.63 R2: 1.85 epi-layer full Q 55Fe # Entries • • • • • • • at n-well partial collection in 6.49 keV Cluster Signal [ADC] Diericks, Meynants, Scheffer (1997) LECC-Heidelberg, 15.06.2004 Norbert Wermes, Bonn 55 Monolithic Pixels / CMOS Active Pixels R&D on CMOS pixels wide spread in HEP: >20 labs (10 IC design groups) Target projects: BELLE upgrade Æ SuperBELLE (< 2008) Univ.Hawaii, KEK, Univ.Tsukuba, Univ.Krakow, Univ.Tokyo, Univ. Nova Gorica STAR upgrade LBNL, IReS (Strasbourg), Univ.Irvine ILC IReS (Strasbourg), DAPNIA (Saclay), LPC (Clermont), LPSC (Grenoble), Univ.Roma-3, Univ.Bergamo, LBNL, RAL, Univ.Hamburg, Univ.Liverpool, Univ.Glasgow, DESY, Univ.Pavia, Univ.Oregon, Univ.Yale, Univ.Como CBM @ GSI IReS (Strasbourg), GSI, Univ.Frankfurt LECC-Heidelberg, 15.06.2004 Norbert Wermes, Bonn 56 CMOS active pixels / R/O & performance row selection → column R/0 „standard“ 3 transistor R/O scheme (e.g. MIMOSA-1, BELLE with pipeline) → upgraded to include amplification, current memory (15 transitors, MIMOSA-7) reset source follower stage • detectors sizes up to 19.4 x 17.4 mm2 (1Mpix) select • smallest pitch: 17 µm • achieved frame speed: 10µs for 132x48 pixels (BELLE) @ 30-50e- noise. from M. Winter LECC-Heidelberg, 15.06.2004 Norbert Wermes, Bonn 57 from D. Contarato LECC-Heidelberg, 15.06.2004 Norbert Wermes, Bonn 58 Monolithic Pixels / achieved features MPV~334 e noise ~10-15eS/N > 20 εmip > 99% 1 Mpix devices in testbeam AMS 0.6µm spatial resolution 1.5 µm (20 µm pix) 5 µm (40 µm pix) (14 µm epi) from M. Winter present favourite AMS 0.35µm OPTO process (10 µm epi) LECC-Heidelberg, 15.06.2004 cluster charge Norbert Wermes, Bonn 59 Monolithic Pixels / radiation tolerance NIEL non-ionizing energy loss 1012 IEL (M. Winter) (ionizing energy loss) -shift of threshold voltages -leakage current in nMOS trans. and intertrans. ILC int. time studies still ongoing ≤1012 neq/cm2 acceptable for AMS 0.35 OPTO LECC-Heidelberg, 15.06.2004 > 1 MRad ok for short integration times Norbert Wermes, Bonn 60 Monolithic Pixels / next steps / improvements improve rad. tolerant designs • guard rings • avoid thick oxide around n-wells thinning Æ 50 µm improve charge collection i.e. increase depleted region • enlarge n-well surface • increase n-well potential • buried electrode e.g. triple n-well technology or photo-FET techniques triple n-well towards large detectors • large reticules • reticle stitching LECC-Heidelberg, 15.06.2004 Norbert Wermes, Bonn 61 Monolithic Pixels / “in early R&D phase” CMOS with SOI technology • high res.(4kΩcm) 300 µm bulk „via“ structure to electr. layer + full Q-collection (drift !) + probably radhard !? - non-Standard technology ~ 1.5 μm ~ 1.0 μm ~ 300 μm prototypes successful SUCIMA (Cracow/Warsaw/Como) produced, but … noise 90Sr 128 x128 matrix 3.0 µm technology 150x150 µm2 pixels 2.4 cm x 2.4 cm first resultsneeds a detector vendor with a CMOS line ! LECC-Heidelberg, 15.06.2004 Norbert Wermes, Bonn 62 Monolithic Pixels / early R&D P. Jarron (CERN) et al. amorphous-SI atop CMOS ICs • a-Si:H layer for Q-coll. (~1000e) - Æ semi-monolithic + standard CMOS technology ! - low signal Æ very low noise required still in its beginnings LBNL (Perez-Mendez), CERN (Jarron) & others a-Si:H Artistic view ~20% H LECC-Heidelberg, 15.06.2004 Norbert Wermes, Bonn 63 DEPFET Pixels LECC-Heidelberg, 15.06.2004 Norbert Wermes, Bonn 64 Monolithic Pixels / DEPFET pixels Potential distribution: source top gate internal gate - bulk p+ n+ n+ p+ p-channel + drain ---- + -+ - potential via axis top-gate / rear contact 1 μm n -+ totally depleted n--substrate internal Gate potential minimum for electrons 50 µm - 300 μm p+ rear contact V [TeSCA-Simulation] (MOS)FET-Transistor integrated in every pixel (first amplification) Local potential minimum (for e- ) under transistor channel Electrons are collected in „internal gate“ and modulate the transistor-current Signal charge removed via clear contact output is a current J. Kemmer und G. Lutz;, NIM A253 (1987) 365 LECC-Heidelberg, 15.06.2004 Norbert Wermes, Bonn 65 Monolithic Pixels / DEPFET pixels Potential distribution: +15V source top gate drain 0V clear 0V bulk internal Gate symmetry axis p+ p n+ n+ n - ----internal gate CLEAR complete ? n Ælow noise CLEAR Æfast pedestal subtraction p+ YES ! (C. Sandow) 50 µm p+ ~1µm n+ - rear contact [TeSCA-Simulation] (MOS)FET-Transistor integrated in every pixel (first amplification) Local potential minimum (for e- ) under transistor channel Electrons are collected in „internal gate“ and modulate the transistor-current Signal charge removed via clear contact output is a current J. Kemmer und G. Lutz;, NIM A253 (1987) 365 LECC-Heidelberg, 15.06.2004 Norbert Wermes, Bonn 66 Monolithic Pixels / DEPFET pixels full (high ohmic) bulk sensitivity Æ large Q for mip very low input capacitance Æ low noise Æ high E-resolution / thin sensors for ILC Æ high speed (>10 MHz / row) Collaboration: Bonn – Mannheim – Munich (MPI,HLL) interested: Aachen, Prague, Valencia, Fermilab, LBL, Cracow, Como LECC-Heidelberg, 15.06.2004 Norbert Wermes, Bonn 67 PXD4 - DEPFET: Two projects on one wafer XEUS (satellite mission) ILC purpose imaging spectroscopy particle tracking sensor size 7.68 x 7.68 cm² 1.3 x 10 cm², 2.2 x 12.5 cm² pixel size 75 µm 25 µm sensor thickness 300 ... 500 µm 50 µm noise 4 el. ENC ~ 100 el. ENC Readout time per row 2.5 µs 20 ns from L. Andricek, P. Lechner PXD4 - DEPFET: Two projects on one wafer gate_2 global source drain_2 common clear drain_1 gate_1 XEUS (satellite mission) ILC purpose imaging spectroscopy particle tracking sensor size 7.68 x 7.68 cm² 1.3 x 10 cm², 2.2 x 12.5 cm² pixel size 75 µm 25 µm sensor thickness 300 ... 500 µm 50 µm noise 4 el. ENC ~ 100 el. ENC Readout time per row 2.5 µs 20 ns from L. Andricek, P. Lechner DEPFET / operation of a DEPFET Matrix • 1 active row DEPFETs are ON R/O Æ CLEAR Æ R/O • all other rows OFF still active for signals Æ low power ! Current Readout chip (read currents, store and subtract pedestals) LECC-Heidelberg, 15.06.2004 - Read cells of a row & store their currents - Clear internal gates of this row completely - Read again (pedestal currents) and subtract by adding currents Norbert Wermes, Bonn 70 DEPFET / operation of a DEPFET Matrix Gate Switcher Hybrid DEPFET Matrix 64x128 pixels, 36 x 28.5µm2 Current Readout CUROII Clear Switcher Testbeam Stack PCB with DEPFET matrix Analog board with ADCs etc. USB based digital interface board from R. Kohrs LECC-Heidelberg, 15.06.2004 Norbert Wermes, Bonn 71 DEPFET / radiation tolerance concern: IEL from beamstrahlung e+e- at ILC 60Co (1.17 MeV and 1.33 MeV) X-Ray tube with Mo target at 30kV pre- vs post-irradiation ∆Not (cm-2) -∆Vth (V) "ON" ~ 4-5 V"OFF" ILC 3.5h 123.5h 293.5h annealing Drain current (µA) Dose (krad) threshold shifts transistor characteristics from L. Andricek LECC-Heidelberg, 15.06.2004 Norbert Wermes, Bonn 72 Performance after irradiation Irradiated double pixel DEPFET o L=7μm, W=25 μm o after 913 krad, 60Co o Vthresh≈-4V, Vgate=-5.3V o Idrain=21 μA o Drain current read out o time cont. shaping τ=6 μs 55Fe Counts/channel o Noise ENC=7.9 e- (rms) at T>23 degC Energy (eV) from L. Andricek DEPFET / testbeam performance correlation telescope ⇔ DEPFET DEPFET data pedestal and common mode corrected Î noise ~ 225 e- resolution multiple scattering dominated ! LECC-Heidelberg, 15.06.2004 Norbert Wermes, Bonn 74 DEPFET / testbeam events 128 Rows 64 columns raw data hit LECC-Heidelberg, 15.06.2004 single hit cluster interpretation δ-electron with perpendicular emission range consistent with measured energy deposit Norbert Wermes, Bonn 75 DEPFET / testbeam results spatial residuals for stiff tracks dominated by multiple scattering (6 ry a in m eli r P GeV e-) Pixel Size 36 µm x 28.5 µm Æ Spatial residuals σres ~ 10 μm pitch Æ Binary position reconstruction ( 12 ) σsp ~ 10.4 µm x 8.3 µm Æ Center of Gravity (S/N ~ 144) expected σsp ~ 2 - 4 µm from L. Reuen LECC-Heidelberg, 15.06.2004 Norbert Wermes, Bonn 76 ILC Module Concept: use good S/N for thinning ! ‘Holes’ in frame save material Chips are thinned to 50 μm, connection via bump bonding Thinned sensor (50 µm) in active area Thick support frame (~300 µm) 50 µm 300 µm Cross section of a module 1st layer module minimal cooling (X0 !!) necessary ~ 5W for 5 layer VTX detector m m .5 15 thinning technology ; proven with active diodes sketch of a(<1nA/cm2) 300µm r= 50 µm 8 Modules in Layer1 from L. Andricek Concluding Remarks ¾ Hybrid Pixels - state of the art for tracking (LHC) Æ imaging (biomed & synchr. light) - SLHC: new materials, 3D ? ¾ (Semi)-Monolithic Pixels - a MUST for the next generation of pixel VTX detectors because of: X0, small pixel size, low power potential - more „mature“ solutions developments - CMOS Active Pixels - DEPFET pixels - very interesting R&D developments: SOI pixels and a-Si:H ¾ ultimate goals (dreams ?) - tracking: full CMOS R/O & logic with charge coll. in high res. bulk - imaging: amorphous high-Z materials (a-Se, PbI2 , HgI2 , PbO) on top of CMOS ICs LECC-Heidelberg, 15.06.2004 Norbert Wermes, Bonn 78 END . LECC-Heidelberg, 15.06.2004 Norbert Wermes, Bonn 79 Backup Transperencies Hybrid Pixels LECC-Heidelberg, 15.06.2004 Norbert Wermes, Bonn 80 # Pixels / chip Pixel area [mμ2] Iana Power/chip Power/pixel Power density [mA] [mA] [mW] [μW] [ mW/cm2 ] ALICE 8192 21’250 150 300 810 99 466 ATLAS 2880 20’000 35 75 190 67 335 CMS 4160 15’000 32 24 121 29 194 LECC-Heidelberg, 15.06.2004 Idig Norbert Wermes, Bonn 81 CMS LECC-Heidelberg, 15.06.2004 Norbert Wermes, Bonn 82 LECC-Heidelberg, 15.06.2004 Norbert Wermes, Bonn 83 LECC-Heidelberg, 15.06.2004 Norbert Wermes, Bonn 84 DEPFET / testbeam results • Purity ≈ 0.97 • Efficiency ≈ 0.99 (6σ seed cut) x 10 S/N = 144 @ 450 µm thickness Î S/N ~ 15 for 50 µm from R. Kohrs LECC-Heidelberg, 15.06.2004 Norbert Wermes, Bonn 85 Because of time walk, if charge is shared, some low pulse heights may be associated to the wrong beam crossing. These lost hits results in degradation of resolution, there is also an issue of intime resolution. A feature of the FE is the ability to duplicate hits with low pulse height (ToT), changing the beam crossing assignement. Therefore the efficiency to collect all hits generated by one track can be greatly improved, giving maximum resolution, with just a moderate increase of occupancy. LECC-Heidelberg, 15.06.2004 efficiency Late hit duplication Cluster efficiency efficiency 1 no hit duplication duplicate ToT 5 0.8 duplicate ToT 10 duplicate ToT 15 0.6 0.4 Efficiency for all hits in the same BCO 0.2 0 0 10 20 30 40 efficiency hit 50 60 70 time (ns) ToT threshold Hits/track No duplication 2.03 ToT<5 2.24 ToT<10 2.71 ToT<15 3.11 Norbert Wermes, Bonn 86 Charge collection in irradiated assemblies Pixel Cluster Charge 200V Unirradiated 0.025 1.1 10 15 2 n eq/cm , 25h@60C 0.02 0.015 400V 0.01 0.005 0 0 600V LECC-Heidelberg, 15.06.2004 10 15 20 25 30 35 40 45 50 Cluster Charge (ke) cluster charge (ke) 700V Vbias 5 mean CCE after 10 yrs LHC ~ 80 % 2 pixels (with LHC type annealing scenario) due to 1. bias grid, 2. trapping Norbert Wermes, Bonn 87 Hybrid Pixels / MPEC / energy windowing multi threshold counting Æ differential E-measurement 2 thresholds + window logic 0.030 tube flux, arbitrary units 0.025 low tube spectrum after 1 cm tissue after 1 cm bone high behind tissue 0.020 behind bone 0.015 0.010 MPEC2.3 0.005 0.000 20 30 40 50 60 energy / keV LECC-Heidelberg, 15.06.2004 70 80 N38-1 Löcker see also N36-106 Edling Norbert Wermes, Bonn 88 Hybrid Pixels / Summary ¾ state of the art technology • large (~m2) rad hard tracking detectors in production 9 • counting pixel detectors for X-ray imaging 9 even at large scale (PILATUS) 9 ¾ issues • • • • complicated assembling (hybridization, R/O flex hybrid) watch yield losses (many production steps) material budget (for tracking detectors) not low (>2% X0) high Z materials (for imaging) not trivial (Q-coll+hybridiz.) ¾ trends • • • • diamond (better than Si ?) smaller pixel cells / interleaved pixels Æ < 5µm resolution MCM-D – hybridization 3D edge active sensors for high rate / large active area LECC-Heidelberg, 15.06.2004 Norbert Wermes, Bonn 89 Indium process (AMS) Rome, 31st of January 2000 Wafer Cleaning Photolithography Process parameters: Plasma activation • Resist Thickness: 15 μm • Pre-bake: 30min @ 80 °C Evaporated Indium • Deposition rate: 0.5 μm/min • Dep. Pressure: 9 x 10 - 7 Torr • Temp. during Dep. < 50 °C Wet Lift off process Anna Maria Fiorello - Research Dept LECC-Heidelberg, 15.06.2004 Norbert Wermes, Bonn 90 Processflow PbSn-Bumping using Electroplating Sputter Etching and Sputtering of the Plating Base / UBM Spin Coating and Printing of Photoresist Resist Stripping and wet Etching of the Plating Base Electroplating of Cu and PbSn Reflow LECC-Heidelberg, 15.06.2004 Norbert Wermes, Bonn 91 PbSn Solder Bump Structure oxidation protection (Au) (100 - 200 nm) (1 - 5 µm) (200 nm) (30 - 200 nm) LECC-Heidelberg, 15.06.2004 Norbert Wermes, Bonn 92 Monolithic Pixels / tracking Î Linear Collider requirements Thin (< 50 µm, 0.1% Xo) Small cells (< 25 µm x 25 µm) Fast (50 MHz/line, 25 kHz/frame ≈2Mpix) Low power (few Watts for full detector) No trigger Options: CCD MAPS HAPS DEPFET Layer Module size No. Of modules I 13 x 100 mm 1 x 8 II 22 x 125 mm 2 x 8 III 22 x 125 mm 2 x 12 IV 22 x 125 mm 2 x 16 V 22 x 125 mm 2 x 20 Total > 500 MPixel (w. 25x25 µm cells) hit rate = 80 hits / mm2 / bunch train due to beamstrahlung e+e- pairs occupancy ~20% Î need tricks : continuous R/O to both sides LECC-Heidelberg, 15.06.2004 Norbert Wermes, Bonn 93 3D LECC-Heidelberg, 15.06.2004 Norbert Wermes, Bonn 94 Speed: planar 3D from S. Parker 4. 4. 4. 1. 3D lateral cell size can be smaller than wafer thickness, so 1. shorter collection distance 2. in 3D, field lines end on cylinders rather than on circles, so 2. 3. most of the signal is induced when the charge is close to the electrode, where the electrode solid angle is large, so planar signals are spread out in time as the charge arrives, and higher average fields for any given maximum field (price: larger electrode capacitance) 3. 3D signals are concentrated in time as the track arrives 4. Landau fluctuations along track arrive sequentially and may cause secondary peaks (see next slide) 4. Landau fluctuations arrive nearly simultaneously 5. if readout has inputs from both n+ and p+ electrodes, 5. drift time corrections can be made 6. for long, narrow pixels and fast electronics, 6. track locations within the pixel can be found LECC-Heidelberg, 15.06.2004 Norbert Wermes, Bonn 95 Keys to the technology S. Parker 1. Plasma etchers can now make deep, near-vertical holes and trenches: a. SF6 in plasma → F, F – → driven onto wafer by E field b. Si + 4F → SiF4 (gas) c. SF6 replaced with C4F8 → CF2 + other fragments which d. form teflon-like wall coat protecting against off-axis F, F – e. repeat (a – d) every 10 – 15 seconds 2. At ~620ºC, ~0.46 Torr, SiH4 gas molecules bounce off the walls many times before they stick, mostly entering and leaving the hole. When they stick, it can be anywhere, so they form a conformal polysilicon coat as the H leaves and the silicon migrates to a lattice site. 3. Gasses such as B2O3, B2H6 (diborane), P2O5, and PH3 (phosphine) can also be deposited in a conformal layer, and make p+ and n+ doped polysilicon. 4. Heating drives the dopants into the single crystal silicon, forming p–n junctions and ohmic contacts there. Large E drift fields can end before the poly, removing that source of large leakage currents. 5. Active edges are made from trench electrodes, capped with an oxide coat. Plasma dicing up to the oxide etch stop makes precise edges. LECC-Heidelberg, 15.06.2004 Norbert Wermes, Bonn 96 MAPS LECC-Heidelberg, 15.06.2004 Norbert Wermes, Bonn 97 CMOS active pixels / towards larger modules 8” wafer (0.25µm technology) large & full reticles seamless reticle stitching ~2x2 cm2 Æ approach 100% active area for modules of order 10cm2 CLRC-RAL LECC-Heidelberg, 15.06.2004 IRES/LEPSI Norbert Wermes, Bonn 98 CMOS pixels / towards more complete charge collection Triple n-well (CLRC&RAL) better field shaping R. Turchetta, Vertex03 photo-FET (IReS&LEPSI) Deptuch/Dulinski IEEE2003 similar to DEPFET pixels charge collected at n-well affects gate voltage of pMOS FET and modulates its current LECC-Heidelberg, 15.06.2004 Norbert Wermes, Bonn 99 CMOS pixels / towards more complete charge collection Triple n-well (CLRC&RAL) better field shaping R. Turchetta, Vertex03 photo-gate (Irvine-LBL) S. Kleinfelder, IEEE2003 LECC-Heidelberg, 15.06.2004 Norbert Wermes, Bonn 100 DEPFET LECC-Heidelberg, 15.06.2004 Norbert Wermes, Bonn 101 DEPFET / three (very different) projects Bioscope a single circular DEPFET Îmedical imaging purpose Fe imaging spectroscopy Xray astronomy 55 tracking particle HEP 64 x 64 pixels 7.68 x 7.68 cm² 1024 x 1024 pixels 1.3 x 10 cm² (x 8) 520 x 4000 pixels (x 8) 4 kpix 1 Mpix 2.1 Mpix (x120)=2.5 x 108 75 µm 25 µm biomed. imaging L = 5 µm,P. W Weilhammer = 40 µm Thursday, detector format 3.2 x 3.2 mm² time-continuous filter, τ9h = 6 µs Next for XEUS development: pixel size µm operation of 64 x 64 50 prototype incl. readout & control ASICs sensor thickness 300 µm 300 ... 500 µm 50 µm matrix noise 65 el. ENC < 4 el. ENC 50-100 el. ENC readout time ... per detector ... per row 1 ms 200 µs 1.2 ms 2.5 µs 50 µs 20 ns status finished Æ 57lp/mm dev. phase dev. phase LECC-Heidelberg, 15.06.2004 Norbert Wermes, Bonn 102 DEPFET thinning Top Wafer Handle <100> Wafer a) oxidation and back side implant of top wafer b) wafer bonding and grinding/polishing of top wafer TESLA-Module („dummy“ sample) 50µm silicon with 350µm (perforated) frame LECC-Heidelberg, 15.06.2004 c) process Î passivation open backside passivation d) anisotropic etching from backside (TMAH) thinned diode structures: leakage current: <1nA /cm2 Norbert Wermes, Bonn 103 DEPFET / Module Concept for LC Auslesechips 520 x 4000 pixel DEPFET-Matrix (25 x 25µm Pixel) Steuerchips Auslesechips LECC-Heidelberg, 15.06.2004 • Sensor area thinned down to 50 µm • Remaining frame for mechanical stability carrying readout and steering chips Norbert Wermes, Bonn 104