M. Gilchriese Local Support R&D Update ATLAS Pixel Upgrade Meeting April 9, 2008 M. Cepeda, S. Dardin, M. Garcia-Sciveres, M. Gilchriese and R. Post LBNL W. Miller and W. Miller iTi C. Daly, B. Kuykendall, H. Lubatti University of Washington

M. Gilchriese 2 Module Dimensions for Studies Have defined multi-chip and single-chip dimensions(two alternatives) as basis for studies. Assumption is that multi-chip used for outer layers and single-chip used for innermost layer(s). Exact break in radius is TBD. See backup for more information Also basis for understanding possible cost reductions in bump deposition and flip-chip (not discussed here) Optimize module size assuming 6” planar sensor wafers

M. Gilchriese 3 Outer Stave Concept Staves for outer barrel layers for multi-chip modules. Based on foam, thin carbon-fiber facings AND flex-cable laminated to stave under modules (bus-cable) Bus-cable routes power, signals, HV to end-of-stave cards. Studies started to see if this works electrically. Bus-cable worst case thermally If bus-cable concept bad…revert to “Type I cables” Modules on both sides (staggered for coverage) Connector from module to connector on bus-cable Dimensions shown for CO2(thicker for C3F8) CARBON FOAM

M. Gilchriese 4 Outer Layer Layout Example

M. Gilchriese 5 Double-Outer Layer Concept Support two outer layers of staves from single shell? Doesn’t look impossible Obviously lots of details…… Composite shell and inner support rings combine assembly into one unit

M. Gilchriese 6 Sensors Heating From Dawson et al. (radiation task force) and temperature parameterization of Unno Figure is for 6000 fb -1 Table except assumes 6000 fb -1. Short strips are at about 30 cm W/cm 2 shown in table

M. Gilchriese 7 FEA Thermal Model Pixel Arrangement –Modules alternate top to bottom, total 5 modules –Take an array of 3 on top to obtain reasonable symmetry in heat spreading for middle module, leaving two on the bottom VG 7 Inputs for thermal runaway calculations

M. Gilchriese 8 Pixel Thermal Model-Baseline Thermal Solution –Carbon foam core K=6W/mK Peak Differential Temperature Center Module –7.63ºC VG 8 Cooling Tube Inner Wall Reference Temperature 0ºC

M. Gilchriese 9 Thermal Performance - I Include detector heating (worst case shown is for total fluence of about Best case shown is for R ~ 16 cm(and 6000 fb -1 ). “Baseline” parameters assumed in thermal model – see backup Looks promising for outer layers. Need higher K’s for (see next page) unless assume colder fluid(<-30) than current C3F8 Remember need to include effect of  T from pressure drops

M. Gilchriese 10 Thermal Performance - II Results below all for fluence and changes in stave-component K values. Need to optimize for Different designs for inner(most) and outer layers Unless CO2. Note CO2 also significantly lower in radiation length Higher K foam Carbon-carbon facings CVD diamond facings No bus cable Combinations….. Note that these studies also apply to 2cm wide stave with single pipe that is more likely at innermost R

M. Gilchriese 11 Pixel Monolithic Structure VG 11 Alternating: Inner and Outer Layer Older module dimensions used for this study

M. Gilchriese 12 Thermal FEA-Based on 0.6W/cm 2 VG 12 Differential from silicon to coolant wall is 10.6˚C. Need improvement to prevent thermal runaway with C3F8….to be studied Foam K=10 W/mK

M. Gilchriese 13 Thermally Conducting Foam Update Obtained additional foam samples – three vendors Made additional small thermal prototypes and measured (see backup for details). Preliminary results. Foam  (g/cc) K(W/m-K)  T max Allcomp 10.18~ 6 We measured ~ 10 Allcomp 20.21Not known~ 9 POCO0.09~ 17(z) ~ 6(x-y) Vendor supplied ~ 11 Koppers * 0.21~ 30(z?) Vendor supplied ~ 8 Average  max at 0.64W/cm 2 on one side * Have 2 other higher density and K samples from Koppers Old prototype, shown last meeting New results, new samples

M. Gilchriese 14 Foam Mechanical Properties Important to measure mechanical properties of foam Being done at University of Washington Allcomp 1 results – see presentation at meeting link at

M. Gilchriese 15 Outlook Optimize thermal performance, radiation length, mechanical properties combining foam(s), facing materials and support. Likely to result in different inner and outer staves. Monolithic still option for innermost layer. Biggest impact on radiation length is choice of coolant… Starting on disks. Layout not as easy as barrel…..particularly in case of two-part system, one inside support tube Overall layout issues and electrical interfaces that drive layout – started some work on this (with UCSC, SLAC, OSU, SMU..) Thermally conducting, carbon foam continues to look promising. Multiple vendors interested (are there more?). Need to optimize and get more samples of what we want. Vendors willing to develop lower density.

M. Gilchriese Backup

M. Gilchriese 2x2 module & stave layouts M. Garcia-Sciveres

M. Gilchriese 18 2 options “Small chip” “Big chip” Boundary between “small” and “big” is determined by the 6” sensor wafer layout that must be compatible with bump bonding (will become clear later) “Small” chip has also a more natural number of rows & columns, but this is probably a minor issue for chip design.

M. Gilchriese 19 Parameters # cols total# rows total# ganged rows # long colsLong col width Small chip Small 2x2 tile um Small active edge 1x1 tile um big chip big 2x2 tile um big active edge 1x1 tile um Normal col. width x row height = 250um x 50um

M. Gilchriese 20 “Small” 4-chip module Active 32.8 x Flex pigtail (connector plugs into page) Pixel orientation Flex down to chip w-bonds 0.2 (vertical inter-chip gap 0.1mm)

M. Gilchriese 21 “Big” 4-chip module Active 35.8 x Flex pigtail (connector plugs into page) Pixel orientation Flex down to chip w-bonds 0.2 (vertical inter-chip gap 0.1mm)

M. Gilchriese 22 Loaded module 20 position connector would be used. Replace dimension by 6.52 glue chips sensor flex connector Reduced scale 1.0 mm stiffener

M. Gilchriese 23 “Small” module outer stave … End of stave card serving 8 modules (half a stave) along Z Can serve one face only (top or bottom) => 4 cards per stave Or can be a wrap-around end of stave card and serve both faces => 2 cards per stave. This way identical staves (including bus cable) design can be used over a wide radial range: 4 cards/stave at lower radius and 2 wrap-around cards per stave at higher radius Module on back 986mm 38.4

M. Gilchriese 24 “Big” module outer stave … End of stave card serving 7 modules (half a stave) along Z (or 8 modules for 1082mm active length) Can serve one face only (top or bottom) => 4 cards per stave Or can be a wrap-around end of stave card and serve both faces => 2 cards per stave. This way identical staves (including bus cable) design can be used over a wide radial range: 4 cards/stave at lower radius and 2 wrap-around cards per stave at higher radius Module on back 946mm 39.9

M. Gilchriese 25 “Small” sensor 6 inch wafer Active area = 7508 mm^2 Sensor tiles shown with darker line Wafer scale flip chip compatible. Chips shown with lighter line. The name “small” 2x2 tile comes from the wafer layout. –A slightly larger chip and therefore larger 2x2 tile is possible, but only 6 such “large” 2x2 tiles will fit on a 6” wafer.

M. Gilchriese 26 “Big” sensor 6 inch wafer Active area = 7539 mm^2 Sensor tiles shown with darker line Wafer scale flip chip compatible. Chips shown with lighter line. OPTION to make 4 6-chip modules per wafer instead of 6 4-chip modules.

M. Gilchriese 27 “Small” single chip module Using same chip as 4-chip module (hence “small”) Active edge sensor 2-side abuttable format active

M. Gilchriese 28 “Big” single chip module Using same chip as 4-chip module (hence “small”) Active edge sensor 2-side abuttable format active

M. Gilchriese Stave Concepts and FEA W. Miller and W. Miller

M. Gilchriese 30 VG 30 Pixel Activities AnalysisAnalysis –Analyze foam structure for inner Pixel Layer steady state chip heating and thermal runaway Thermal runaway evaluation covers different foam and facing thermal conductivitiesThermal runaway evaluation covers different foam and facing thermal conductivities Design LayoutDesign Layout – Preliminary stages of evaluating packaging for layers at 16cm and 21cm radius TestingTesting –Thermal solutions to compare with LBNL stave/carbon foam core thermal tests Evaluation embraces several foam core thermal conductivitiesEvaluation embraces several foam core thermal conductivities

M. Gilchriese 31 Pixel Stave Structure Stave Analysis- 1 meter length –In the near future an effort will be underway to assess structural aspects of stave concept for pixels For now focusing on thermal effects –Pixel sensor is mm by mm –Pixel chip footprint, 4 total, is 38.4 mm by 38.4mm –Assumed pixel heat load is 0.6W/cm 2 –Small diameter cooling tube (presumes CO 2 ) Steps in process –1 st Order thermal analysis of sandwich structure (conductive carbon foam core) –Several solutions made for thermal runaway Looks workable without CVD diamond sandwich facings –This model is still being evaluated

M. Gilchriese 32 Basic Model Parameters-Baseline Core –Carbon foam, 6 W/mK Facing –Resin Composite, 0.14mm thick, 110, 1, 110 (X,Y,Z) W/mK Cable –Includes adhesive for bonding to chips and from cable to composite facing –2mils Al and 0.7mils of copper, plus adhesives, total compressed thickness=114microns –Calculated: K t =0.38W/mK and K (in-plane)=83W/mK Sensor Chip heat 0.6W/cm 2

M. Gilchriese 33 FEA Thermal Model Pixel Arrangement –Modules alternate top to bottom, total 5 modules –Take an array of 3 on top to obtain reasonable symmetry in heat spreading for middle module, leaving two on the bottom VG 33 Inputs for thermal runaway calculations

M. Gilchriese 34 Pixel Thermal Model-Baseline Thermal Solution –Carbon foam core K=6W/mK Peak Differential Temperature Center Module –7.63ºC VG 34 Cooling Tube Inner Wall Reference Temperature 0ºC

M. Gilchriese 35 Pixel Thermal Model Thermal Solution –Carbon foam core K=100W/mK Peak Differential Temperature Center Module –4.05ºC VG 35 Very High Foam Conductivity alters peak differential by 3.58ºC

M. Gilchriese 36 Thermal Runaway-Baseline 1*10 16 fluence makes -25ºC impractical without design changes to sandwich material maheup

M. Gilchriese 37 Thermal Runaway with Possible Mod’s Results show increasing conductivity of foam and facing thermal conductivity improve situation noticeably Thermal solution with CVD diamond facing still under evaluation, all indications is that it may be overkill Fluence of 1*10 16

M. Gilchriese 38 Design Layouts-In Process Beginnings of 210mm and 160mm layers Composite shell and inner support rings combine assembly into one unit

M. Gilchriese 39 Design Layouts-In Process Space between adjacent staves is very tight, suggesting that the stave support from the rings may best engage area between module dead-spaces

M. Gilchriese 40 Design Layouts-In Process First option for 1m length is three support rings, one in middle which will provide the Z-restraint and the two at the ends reacting out gravitational effects, but allowing slip in Z Ring locations

M. Gilchriese 41 Carbon Foam Thermal Tests Following slides provide background on carbon foam thermal conductivityFollowing slides provide background on carbon foam thermal conductivity VG 41

M. Gilchriese 42 VG 42 FEA Model Primary ObjectivePrimary Objective –Compare FEA results with LBNL thermal tests of foam core structures Difficulty lies in assigning material propertiesDifficulty lies in assigning material properties –There are four solids, with three thermal interfaces on each side of the mid-plane Thermal interface thermal resistance becomes an assumption, as well as the thicknessThermal interface thermal resistance becomes an assumption, as well as the thickness –Water coolant Flow results in turbulent flow and very high convection coefficient, less problematic than thermal interface resistanceFlow results in turbulent flow and very high convection coefficient, less problematic than thermal interface resistance Expect small variations in coolant temperature from test to testExpect small variations in coolant temperature from test to test

M. Gilchriese 43 VG 43 Solution With FEA Model Material PropertiesMaterial Properties –Heater heat loads, 8.38W –Silicon heater, 148 W/mK, 0.28mm thick –Silicon heater adhesive, SE4445, 0.6 W/mK, 0.004in thick, two places –YSH70 open cloth fabric, one layer, 0.6 W/mK, 0.14mm –YSH70 adhesive, 1.55 W/mK, 0.002in –Foam properties varied, from 6 to 30 W/mK –Al cooling tube, 180 W/mK, 2.8mm OD and 2.19mm ID –Water, convective film coefficient, 66,000 W/m 2 K, 1.0L/min Set 20.25ºC on inner tube wallSet 20.25ºC on inner tube wall –K13D2U facing, 1 W/mK, 0.28mm thick –K13D2U adhesive, 1.55 W/mK, 0.002in thick

M. Gilchriese 44 VG 44 Pixel Prototype Components Tube with CGL7018 YSH-70 and K13D2U glued to foam Tube in foam with CGL7018

M. Gilchriese 45 VG 45 LBNL Thermal Test Set-Up Silicon heater

M. Gilchriese 46 Thermal Solutions for Single Tube Tests Double heaterSingle heater

M. Gilchriese Prototype Details

M. Gilchriese 48 Old Results Note if CO2 used as coolant then reference temperature could be about -30C. Thus delta T of 10 => T of -20C. FE-I4 goal FE-I3 normal Max. spec  Includes sensors & power conv. But not cables.  

M. Gilchriese 49 Old Results Table All relative to 20C water temperature, would be slightly lower if referenced to power off temperature.

M. Gilchriese 50 New Prototypes Identical width, thickness and adhesives to older prototype (Allcomp 1) but shorter in length (7.4 cm). YSH-70 facings on both sides. Heater only on one side. Compare at 0.64 W/cm 2 IR and water flow same as older prototoype(1.0 l/min)

M. Gilchriese 51 IR Results Example IR photo(Koppers) Average T in boxes used Small difference between power off  T and water  T Two different values for Allcomp 2 in table below. One(29 twice) I ignore apparent hot spot on heater. Other entry I don’t.