1. IEEE Nuclear Science Symposium Nov. 8 2001, San Diego, CA Professor Priscilla Cushman University of Minnesota Problems and Solutions in high-rate multi-channel Hybrid Photodiode design HPD’s for the CMS Hadronic Calorimeter Professor Priscilla Cushman University of Minnesota The US-CMS HCAL Collaboration Fermilab Florida State Purdue Notre Dame University of Illinois (Chicago) University of Mississippi University of Maryland Rochester University of Minnesota

2. IEEE Nuclear Science Symposium Nov. 8 2001, San Diego, CA Professor Priscilla Cushman University of Minnesota HCAL The CMS Tile/Fiber Hadronic Calorimeter

3. IEEE Nuclear Science Symposium Nov. 8 2001, San Diego, CA Professor Priscilla Cushman University of Minnesota Reading out the Towers of Tiles with WLS fiber and HPD’s

4. IEEE Nuclear Science Symposium Nov. 8 2001, San Diego, CA Professor Priscilla Cushman University of Minnesota Optical Decoder Unit HPD mount aligned to cookie and plate Fiber Optic Cables attach to a patch panel Optical decoding from layer into tower bundles occurs at the readout boxes.

5. IEEE Nuclear Science Symposium Nov. 8 2001, San Diego, CA Professor Priscilla Cushman University of Minnesota Stringent Photodetector Requirements Magnetic Field of 4 Tesla Linear Response from MIP to 3 TeV Shower DC Calibration to 2% using Radioactive Source Integrated Neutron Dose of up to 5 x 10 10 n/cm 2 Integrated Output Charge up to 3 Coulombs Use same system in HO for ease of integration Outer Barrel and Endcap have relaxed constraints tag muons (10 p.e./MIP) and measure shower tails many channels => low cost fringe field

6. IEEE Nuclear Science Symposium Nov. 8 2001, San Diego, CA Professor Priscilla Cushman University of Minnesota The Hybrid Photodiode DEP Tube Fabrication by Delft Electronic Products (Netherlands) Subcontracts: Canberra (Belgium) diodes Schott Glass (USA) fiber optic windows Kyocera (Japan) vacuum feedthru/ceramic carrier 19 x 5.4mm73 x 2.68mm CMS diode design 12 kV across 3.3 mm gap with V th Gain of 2500 Silicon PIN diode array, T-type. Operated at 80 V reverse bias Thin (200  m) with 100 V reverse bias for fast charge (holes) collection Aluminized surface and AR Coating

7. IEEE Nuclear Science Symposium Nov. 8 2001, San Diego, CA Professor Priscilla Cushman University of Minnesota How They Work PIN Diode array Ceramic feedthrough Fiber-Optic Window Photocathode Electrons from the photocathode are accelerated in a high electric field and stop in the diode where they generate electron hole pairs. Detect current as holes move across the depletion region in the back-illuminated version. e 4 16 kV Gain 0 0 4000 8

8. IEEE Nuclear Science Symposium Nov. 8 2001, San Diego, CA Professor Priscilla Cushman University of Minnesota APD vs HPD decision (1995-6 CERN test beams) System Integration Issues NIM A387 (1997) 107 Signal/Noise at low light-levels AC: 1 MIP=7-10 pe DC: Radioactive Source Validation B-field Studies at 4-5 T NIM A418 (1998) 300 Radiation Damage at Oak Ridge NIM A411 (1998) 304 Extensive Bench Studies Project Approval and CMS-specific design Negotiate specifications and price Acquire prototypes and testNIM A442 (2000) 289 From Validation to Quality Assurance and Yield This has been LONGER than we expected ! Iterate Anatomy of a Decision

9. IEEE Nuclear Science Symposium Nov. 8 2001, San Diego, CA Professor Priscilla Cushman University of Minnesota Magnetic Field Issues Performance measured at 4 Tesla Doesn’t break Image shift = gap * tan  Small gain shifts (angular effects and backscattter) Align tube axis parallel to field Field locally uniform: ~ 5 o (6 o ) inclination in HB (HE) Minimum gap (eases tolerance) vs HV (maintain gain) Maintain sufficient space between fiber bundles Mechanically robust cookies Maximize photocathode active area Pixel position must be measured (and aligned) to 50  m

10. IEEE Nuclear Science Symposium Nov. 8 2001, San Diego, CA Professor Priscilla Cushman University of Minnesota Larger active area: Less room for HV connection, possible field distortions Minimize gap: Improve tube components This has been a development project This has been a development project (the tube) HV coax cable : Reliability and compatibility with RBX RBX mountings and cookies must be rubber insulated Gold-plated pins: Enables us to use ZIF sockets - questions of gold diffusion FIBERS RBX HPD RBX Cookie HV Cable Electronics Interface Board

11. IEEE Nuclear Science Symposium Nov. 8 2001, San Diego, CA Professor Priscilla Cushman University of Minnesota HV Discharge to HPD Mounting Current across mount at 12 kV with normal RBX configuration nAmps Lifetime Setup modified for HV Monitor Trigger: 1 nA for spikes + 1.5 minute interval Time (hrs) After silicon rubber potting, at 15 kV nAmps Time (hrs)

12. IEEE Nuclear Science Symposium Nov. 8 2001, San Diego, CA Professor Priscilla Cushman University of Minnesota Richardson-Dushman Equation J th = A o T 2 exp(-  eff /kT)  eff = 1.09 for this fit. Photocathode has a typical e -   dependence on Temperature Keep red sensitivity low. Fitted  eff ranged from 1 - 1.33 Dark counts at 25 o C are 50 Hz – 1 kHz Temperature ( o C)

13. IEEE Nuclear Science Symposium Nov. 8 2001, San Diego, CA Professor Priscilla Cushman University of Minnesota Custom Pixel Design: 19 ch (towers) and 73 ch (short stacks) This has been a development project This has been a development project ( the diode) 2 side-contacts (100 nm thick Al) Bump-bonded vacuum feedthru n++ contact n++ p+ n+ bulk (200  m thick) AR (16 nm sputtered Si) Metal (25 nm AL) Barrier (25 nm SiO 2 ) Higher pin-out density: wire-bonds =>glass feedthrus => ceramic from Kyocera Alignment to 50  m: manufacturer tolerances tightened, new measurement procedures Improved rise time: Thinner silicon: 200  m replaces 300  m Guard ring and drain structure: lower leakage current and better uniformity for edge pixels Lower depletion voltage and better control of process: higher breakdown voltage Surface aluminization and edge traces: Reduce negative crosstalk: 300  /sq => 1.7  /sq Anti-reflective coating: Reduce positive crosstalk from reflected light

14. IEEE Nuclear Science Symposium Nov. 8 2001, San Diego, CA Professor Priscilla Cushman University of Minnesota Capacitance as a function of reverse bias voltage 33 pF 18 pF Depletion at 43 V and 25 V C 73 = 5 pF/pixel C 19 = 20 pF/pixel C(interface card and leads) = 13 pF

15. IEEE Nuclear Science Symposium Nov. 8 2001, San Diego, CA Professor Priscilla Cushman University of Minnesota pe’s Internal Electric Field in the bulk n-type silicon Output current and pulse width can be calculated with this simple model from t = 0 to t =  V=V b V=0 n++ p+ n+ E(0) =V b -V d d x E x=dx=0 E(d) = 2V d + E(0) = V b +V d d d h h h h h

16. IEEE Nuclear Science Symposium Nov. 8 2001, San Diego, CA Professor Priscilla Cushman University of Minnesota Drift time of holes translates into pulse width  Simple form for over-depletion matches data Pulse Width From fit:Consistency check: Hole mobility in silicon is  = 450 cm 2 /Vs

17. IEEE Nuclear Science Symposium Nov. 8 2001, San Diego, CA Professor Priscilla Cushman University of Minnesota 300  m thick200  m thick Pulse width can be shortened by reducing wafer thickness d or by increasing bias voltage, V b Drift time is approximately given by and the shape of the plateau mirrors the internal electric field

18. IEEE Nuclear Science Symposium Nov. 8 2001, San Diego, CA Professor Priscilla Cushman University of Minnesota 200 100 66 50 40 33 28.5 25 Volts Bias Voltage For higher depletion (lower ohmic silicon) V b -V d ~ V b is not true Fit data to model early (5 k  -cm silicon) diodes 12 k  -cm diodes Operating bias voltage = 80 V We now SPECIFY > 8 k  -cm Pulse Width  50 ns 40 ns 30 ns 20 ns 10 ns 0 1/V b

19. IEEE Nuclear Science Symposium Nov. 8 2001, San Diego, CA Professor Priscilla Cushman University of Minnesota No evidence of breakdown, even at 500 volts!

20. IEEE Nuclear Science Symposium Nov. 8 2001, San Diego, CA Professor Priscilla Cushman University of Minnesota Pulse Width in nsec Flatter plateau and higher breakdown for new diodes R0003241 (200 micron, 73-channel new-style diode) Reminder: on the same scale

21. IEEE Nuclear Science Symposium Nov. 8 2001, San Diego, CA Professor Priscilla Cushman University of Minnesota AC Crosstalk eliminated by Aluminization 3334353637 Pixels in center row Positive crossstalk now observed !

22. IEEE Nuclear Science Symposium Nov. 8 2001, San Diego, CA Professor Priscilla Cushman University of Minnesota Backscatter crosstalk for a 19-channel Aluminized HPD Move fiberRead out nearest neighbor Convolution of hexagonal pixel shape with backscatter radial distribution d

23. IEEE Nuclear Science Symposium Nov. 8 2001, San Diego, CA Professor Priscilla Cushman University of Minnesota 0 o 45 o 90 o Initial scattering angle of backscattered electron Radial Distance from impact point (mm) 76543217654321 Ballistic model of Backscattering. 10 keV electrons

24. IEEE Nuclear Science Symposium Nov. 8 2001, San Diego, CA Professor Priscilla Cushman University of Minnesota Trajectories of Backscattered Electrons B = 0 T  = 45 o B = 0.15 T  = 45 o B = 4 T  = 75 o Radial distance Radial distance is a minimum here

25. IEEE Nuclear Science Symposium Nov. 8 2001, San Diego, CA Professor Priscilla Cushman University of Minnesota Number of backscattered e’s as a function of radial distance from impact point 0 1 2 3 4 5 6 7 mm 0 0.5 1 1.5 2 2.5 3 3.5 4 mm 0 0.5 1 1.5 2 2.5 3 mm 0.025.05.075.1.125.15 mm B = 0 TB = 0.15 T B = 4 TB = 0.2 T

26. IEEE Nuclear Science Symposium Nov. 8 2001, San Diego, CA Professor Priscilla Cushman University of Minnesota Positive crosstalk from backscatter can be removed by B-field 73-ch tube Total DC Crosstalk ( 2.7 mm pixels) B=0 B=1.5 T Bare silicon 18% 7% Aluminized 29% 16% Al with AR 12.4% 4.3% STILL some positive crosstalk and Al is worse than silicon 33-41 1-2 72-73 Pixel number

27. IEEE Nuclear Science Symposium Nov. 8 2001, San Diego, CA Professor Priscilla Cushman University of Minnesota photoelectrons Light Re-emitted photoelectrons APD views reflected light pe backscatter focussed by B Light injected thru fiber Test Confirms Reflected Light DIODE ARRAY FIBER OPTIC & PHOTOCATHODE

28. IEEE Nuclear Science Symposium Nov. 8 2001, San Diego, CA Professor Priscilla Cushman University of Minnesota Compare residual HPD crosstalk in B-Field with APD measurement of optical reflection

29. IEEE Nuclear Science Symposium Nov. 8 2001, San Diego, CA Professor Priscilla Cushman University of Minnesota Study Problem at Minnesota, then export technology to DEP IMD - optical modeling package for multilayer structures (by David L. Windt, http://cletus.phys.columbia.edu/windt/idl ) Ag oxidizes quickly Au not available Model Data monochrometer PIN diode Samples: glass slides with various coatings (PECVD) 10 nm Ag 120 nm SiO 2 8 nm A-Si 15 nm Ag 50 nm SiO 2 16 nm A-Si 25 nm Al 25 nm SiO 2 Some Options DEP makes samples on old diodes using sputtering

30. IEEE Nuclear Science Symposium Nov. 8 2001, San Diego, CA Professor Priscilla Cushman University of Minnesota 5.4 min 10.8 nm 20 min 52.1 nm 30 min 76.7 nm 60 min 153.5 nm Comparison of Model vs Data calibrates PE-CVD Process Comparison of Model vs Data calibrates PE-CVD Process a-Si:H (SiH 4 gas) deposited on glass slides

31. IEEE Nuclear Science Symposium Nov. 8 2001, San Diego, CA Professor Priscilla Cushman University of Minnesota Glass slides + 25 nm Al + 14 nm a-Si:H

32. IEEE Nuclear Science Symposium Nov. 8 2001, San Diego, CA Professor Priscilla Cushman University of Minnesota Minnesota test slides 14 nm a-SiH 25 nm Al 25 nm SiO 2 First DEP attempt at 16 nm a-Si MC for 10 nm We tell them to add another 6 nm MC for 16 nm Reflectance Studies Angular Dependence of test slides

33. IEEE Nuclear Science Symposium Nov. 8 2001, San Diego, CA Professor Priscilla Cushman University of Minnesota Specifications are finalized in Contract Quality Assurance protocol detailed

34. IEEE Nuclear Science Symposium Nov. 8 2001, San Diego, CA Professor Priscilla Cushman University of Minnesota Specifications are finalized in Contract...continued

35. IEEE Nuclear Science Symposium Nov. 8 2001, San Diego, CA Professor Priscilla Cushman University of Minnesota Quality Assurance failpass Return to DEP Bake-out at 13 kV for 2 weeks Evaluate 500 tubes, automated procedure, complete web-accessible database Leakage current for each pixel and guard ring @ 80VHV gain, reverse bias curve DC Station Alignment measurements for 50 micron tolerance Machine custom ring crosstalk checks alignment 2-D response scans (10kV, 80 V) pe spectra, AC xtalk capacitance vs bias To FNAL for installation in readout boxes AC Station Lifetime: Q, Cf 252, HV High Rate and B-Field tests subset

36. IEEE Nuclear Science Symposium Nov. 8 2001, San Diego, CA Professor Priscilla Cushman University of Minnesota Precision Registration Test fixture = Standard Mount + metal plate with 3 alignment holes Scan to find centroid of alignment holes Same scan finds pixel intersections above and below metal piece by iterative sector equalization Machine shop uses measured  x,  y,  to produce Custom Ring Stabilized light source HPD Mount Scanning Table Integrating sphere Focussing optics Green filter Metal alignment plate

37. IEEE Nuclear Science Symposium Nov. 8 2001, San Diego, CA Professor Priscilla Cushman University of Minnesota FIBERS Plate HPD Alignment Pins Plate Cookie Ring HV Cable Electronics Interface Board Optical Decoding Unit Each ring is registered to its HPD via alignment pins Plate and Cookies are universal Precision Registration (in assembly)

38. IEEE Nuclear Science Symposium Nov. 8 2001, San Diego, CA Professor Priscilla Cushman University of Minnesota Universal Cookie + Plate 73-channel (HB)19-channel (HB right)

39. IEEE Nuclear Science Symposium Nov. 8 2001, San Diego, CA Professor Priscilla Cushman University of Minnesota HPD alignment after correction for 8 tubes

40. IEEE Nuclear Science Symposium Nov. 8 2001, San Diego, CA Professor Priscilla Cushman University of Minnesota Samples from DC database Pixel Number Dark Current 1 8.6827e-009 2 9.2811e-009 3 2.4709e-008 4 8.6967e-009 5 8.995e-009 6 2.2894e-008 7 8.2391e-009 8 2.1001e-008 9 9.1783e-009 10 2.2305e-008 11 1.36793e-008 12 8.7693e-009 13 2.0977e-008 14 3.0818e-008 15 1.78352e-008 16 7.1035e-009 17 7.2061e-009 18 8.2694e-009 19 7.2356e-009 Total Current 2.6605e-007

41. IEEE Nuclear Science Symposium Nov. 8 2001, San Diego, CA Professor Priscilla Cushman University of Minnesota AC Station: Viking chip serial readout at 10 MHz HPD + interface card AC-Coupling 2 chip 128 channel PA Repeater card Laptop with ADC card 128 Multiplexer Shaper Sample & hold 10 MHz readout

42. IEEE Nuclear Science Symposium Nov. 8 2001, San Diego, CA Professor Priscilla Cushman University of Minnesota Individual spectra for each pixel from AC Station 19-channel tube at low light levels

43. IEEE Nuclear Science Symposium Nov. 8 2001, San Diego, CA Professor Priscilla Cushman University of Minnesota Lifetime Issues PIN reference diodes 1.0 mm diam. WLS fibers Blue LED’s Pixel 1 HPD Pixel 2 Lifetime Monitoring Stations monitor current (PIN diodes, HPD) and temperature Radiation Damage: (10 CMS years = 5 x 10 10 n/cm 2 in worst region) Expose samples to Cf 252 Oak Ridge: Early HPD version to 10 13 n/cm 2 in 1997 tests. Minnesota: Low flux drawer instrumented Aluminized new HPD to >10 11 n/cm 2 in 2001 Integrated Charge: (10 CMS years = 3 C over 25.6 mm 2 pixel at high  Expose to accelerated rate plus control pixel at CMS rate. Surface scans done before and after exposure distinguish between photocathode degradation and silicon damage.

44. IEEE Nuclear Science Symposium Nov. 8 2001, San Diego, CA Professor Priscilla Cushman University of Minnesota Red light with HV=0 Scans only silicon Green light with HV = 8 kV Scans total tube response Concentrated light in a small fiber will damage photocathode This is far beyond the CMS rate which allows for photocathode self-annealing

45. IEEE Nuclear Science Symposium Nov. 8 2001, San Diego, CA Professor Priscilla Cushman University of Minnesota 6 CMS months for high  tower at expected CMS rate 6 CMS years for high  tower at 13 x expected rate Accelerated aging tests Accelerated aging tests: Red (upper curves) = 73-ch aluminized HPD Blue (lower curves) = old tube with poor potting Lifetime setup traps sparking by triggering on current spikes from High Voltage Supply All curves normalized to reference diode and corrected for temperature shifts

46. IEEE Nuclear Science Symposium Nov. 8 2001, San Diego, CA Professor Priscilla Cushman University of Minnesota Irradiate 73-channel aluminized HPD - leakage current increases Leakage current rising at 46 pA/hr PIN diode => Injected light is constant HPD (light-dark) hours Flux = 7 x 10 9 n/(cm 2 hr) MeV neutrons from Cf 252 Day 1Day 2Day 3Day 4 Integrated dose is geometrically equivalent to 4.5 x 10 11 n/cm 2 head on 10 CMS years

47. IEEE Nuclear Science Symposium Nov. 8 2001, San Diego, CA Professor Priscilla Cushman University of Minnesota Conclusions 5 years ago, no existing technology could satisfy our specifications. Development project was initiated with one Company - DEP with backup plans which included Hamamatsu and Litton Rigorous evaluation must include accelerated aging and test beams and enough prototype detectors to understand the yield. The anticipated problems are not the ones that really bite you. This takes time! In the last year, the HPD subsystem has approached “critical path” in the CMS Project. Final Result : CMS HCAL gets what it needs (so we can find the Higgs) and a better product is offered to the general public.