Mechanical and Thermal Management for ATLAS Upgrade Silicon Tracking System

1. Mechanical and Thermal Management for ATLAS Upgrade Silicon Tracking System W. O. Miller (iTi) Carl Haber (LBNL), Gil Gilchriese (LBNL)

2. ATLAS Silicon Tracker Upgrade • Stave-Type Tracking Detector---- Considerations • Dimensional (Static) stability • Out-of-plane distortion from gravity to be <50 to 60microns • Thermal distortions from sub-cooling detector to be <25microns • Cooling • Possibly based on ATLAS two-phase flow, with -20 to -25ºC at detector surface • Or possibly high pressure CO2 at -35ºC inlet (1000psi) • An approach that assures that thermal runaway is not a problem • Places cooling directly beneath chip heat load • Radiation length-minimized, consistent with achieving mechanical objectives • Ease of Assembly and maintenance (when required) • Less time consuming • Precision module placement and bonding at earliest construction level • Facilitates testing and placement calibration • Added benefit in that provides potential for wider participation from institutions in building staves and detector assembly VG 2

3. ATLAS Tracker: Stave (Upgrade) • Analysis scope • Stave Stiffness, evaluate ability to resist gravity sag, critical at upper and lower Φ • Evaluate stave thermal/mechanical aspects • Direct heat transfer to evaporative cooling tubes • Thermal strains from thermal gradient from chips to tubes • Thermal strains from cool-down, room temperature to initial state (minus 25ºC or minus 35ºC) • Coolant tube sizing spaces the facing separation • Analysis guided by realistic constraints on radiation length • Radiation length driven design as always, issues focus on: • Composite material properties • High modulus, optimized fiber orientation for maximum axial stiffness • High conductivity fiber system for heat transport • Issue to be addressed is CTE mismatch in the assembly • Al cooling tube • Stave length, gravity sag, bounded between fixed end and simple supports • ~60μm for fixed to ~320 μm for simple • Desire limit of <60microns VG 3

4. Stave Stiffness( beware discussion covers different geometries)

5. Stave Static Stiffness Discussion • Results of FEA solutions • Stiffness---solution for purely horizontal and purely vertical • Model comprises • Composite sandwich concept----embedded cooling tube • Composite Facings and side strips----primary stiffness element • Fiber choice---K13D2U, very high modulus (120Msi) with thermal conductivity of 800W/mK • Fiber orientation---- 4 to 1 lay up, preferentially aligned along the stave axis • Honeycomb----- graphite fiber honeycomb with 3/16in cell size • Al End caps and stainless steel pins VG 5

6. Stave Module Layout-Staggered Modules Original model Modules staggered front to back ~100W Thermal Load width 6.4cm Original 4.6mm flat tube Physical length of stave ~99.4cm Port card VG 6

7. Stave Module Layout-Double-Sided Modules More Recent Model • Silicon Modules-12.4cm long with 4 sets of chips per strip detector • Chip heat load – 192W • Module heat load -4W/module or 64W/stave 1064mm 71.5mm VG 7

8. Prototype Stave- Tooling • Assembly tooling set to longer and wider stave dimensions • Prototype coolant tube initially considered was 7.3mm flat tube (availability), but now proceeding with 4.7mm height and 14mil wall (purchase new tube extrusion) 1064mm 71.5mm VG 8

9. Consequences of Staggered vs Double-Sided • Silicon Module Lay-out • Interest shown in double-sided geometry • Analysis and lay-outs adapted to provide information for both • Most of the FEA work done on original layout with fewer modules • Adapted the results by adding analytical solutions • Compared analytical to FEA to form basis for making extrapolations • Separation between facings • 4.6mm, 5.88mm, and 7.3mm • Now focusing on 4.7mm • Other factors • Thermal heat load • Coolant pressure • C3F8 and CO2, substantial difference in pressure • Necessitated adapting design to round tube for stress considerations VG 9

10. Stave Support-via Alignment Pins • Stave End Cap • Two pins at each end for stave support and alignment • Pin engagement in stave and end support disk are important to stave stiffness • Sag will affected significantly by the end plate design Strip detector Strip detector ~1m length ~1m length pins Port card Electronics chips VG 10

11. Stave Configuration • Composite Sandwich • Facings high conductivity, high modulus K13D2U in 4:1 orientation (modulus 1.77 times steel) • Honeycomb core fills void not occupied by cooling tube • Cooling: Evaporative • Cooling tube: U-shaped, allowing for entrance and exit at one end • Tube flatten slightly to improve thermal contact • Module Mounting on both composite facings: balancing CTE mismatch to reasonable level • Detector modules mounted on both sides, in either alternating pattern or entirely symmetrical • Hybrids mounted between detector modules More recently modules on both sides What key factors describe stiffness for this sandwich? VG 11

12. Stave Parameters Affecting Stiffness • Stave gravity sag (bending) results from uniform length-wise loading • W=uniform loading, L=unsupported length, C varies from 1 to 5 based on end fixity, b=width of sandwich (w is generally a function of b so this essentially drops out) • D=flexural rigidity, E=facing Young’s modulus, t=facing thickness, h=overall sandwich height, c=core height, ν=facing Poisson's ratio • Where do we gain most? • Minimizing unsupported stave length, next • Maximizing end support rigidity, next • Separation between facings, finally • Facing modulus and facing thickness • What difficulties do we encounter? • Achieving stave lengths >1m by limiting δ < 60microns • Cooling both composite faces with single cooling tube, limits c • Radiation length limits, t • Quasi-isotropic facings, would place a lower limit on E VG 12

13. Coolant Mass Used density of two phase fluid. Mean density is 60kg/m3, whereas liquid density is 1660kg/m3 Round circular tube in half length model (5.88mm) Sandwich Core (shear deflection) Varied core shear modulus, reflected in density change to material 66 to 210kg/m3, CVD carbon foam 56 and 110kg/m3, honeycomb Sandwich core height 4.6, 5.88, 7.33, and now 4.7mm Size for prototype stave initially driven by tube availability---8mm extrusion, but now we are proceeding with a new extrusion Full Length (96cm-4.6mm flat tube) model with wafers, hybrids and cable as dead weight 0.173in diameter support pins Clamped support pins, vertical DOF only Core thickness 4.6mm Half length model with wafers, hybrids and cable as dead weight 0.173in support pins ½ length model used to add structural coupling of wafers, and hybrids Core thickness 5.88mm Model of pins and end cap alone with stave weight imposed Two pin diameters (0.173in and 0.25in) Stave Sag FEA Models (Three) VG 13

14. FEA Sandwich Core Summary-Staggered VG 14

15. Stave FEA Results-Core Shear • Varied core shear modulus • Looked at carbon foam versus honeycomb • Results did not come out entirely as expected • Example: • Increased shear modulus led to increased sag • Increased shear modulus with carbon foam comes at expense of increased core density, this seemed to be over-riding factor • Suspicion: coolant tube with direct coupling to facing was a major contributor to shear load transfer • Approach • Estimate degree of sag from core shear • Uncouple tube by reducing the modulus VG 15

16. c h Idealized Sandwich Stiffness 1.6microns • Deflection from Uniform Loading • Bending Deflection • Core Shear Deflection • End Cap Strain (mounting pins): somewhere between simple and fixed pins Sandwich side L-unsupported length w-load per unit length End cap strain Sandwich dimensions G-core shear modulus b t E-facing modulus C=1 for fixed and C=5 for simple VG 16

17. Establish Cooling Tube Interaction • 96cm Model of Stave • Use simple edge supports, K13D2U 4/1 • Apply force at quarter points, • Extract deflection at Δ4 and Δ2, quarter point and mid-span • Use relationship to solve for core shear modulus • Result for 4.6mm core with Al tubes • ~128 MPa calculated versus 26.9 MPa input for virgin foam • Tubes contribute most of the shear stiffness, except at very high foam densities P/2 P/2 Division between bending and shear, based on estimate of core shear of 128MPa Using sandwich relationships for fixed ends Δbending est=35μm Δcore shear est=8.3 μm Combined Δ=43.3microns (FEA 53.7 μm) VG 17

18. Removed Cooling Tube Interaction • Cooling tube as a shear component is affected by compliance of thermally enhanced adhesive bond (CGL7018) • FEA model modification placed reliance for shear transfer on the composite fiber honeycomb, 626MPa • Increase in sag by small amount (~4microns for HC only) • Experimental tests with stave prototype hopefully will reveal extent of contribution from coolant tube on shear • If tube contributes significantly to shear load transfer, what will be the benefit? • Slight increase in stiffness predicted, but if higher than expected, may be possible to reduce the number of uni-tape layers in the composite facing • Slight benefit in radiation length in using different core material VG 18

19. Modification to End Cap Configuration • Purpose • Lengthen pin engagement with support plate • Increase internal bonding contact with the composite facings • Add tooling balls for assembly inspection Steel pins 0.173in dia Alignment ball Shell-like to minimize mass 1/4in bonding surface Cooling tube penetrations VG 19

20. Stave Radiation Length Estimate Sag Est. 30 µms (extrapolated) ~ prototype 54 µms (FEA) VG 20

21. Vertical Orientation- Version A Stave • Maximum sag 3.5microns • Al end caps • SS pins, 0.173in dia. • 4.6mm separation VG 21

22. Horizontal Orientation- Version A Stave • Maximum sag 55.5 microns • Al end caps • SS pins, 0.173in dia. • 4.6mm separation VG 22

23. 2m Long Stave • Sag • For fixed geometry previously discussed, sag from bending alone would increase by L to the 4th • 50microns would increase to 800microns, without factoring in potential increase in core shear • Increased Sag can be mitigated by increasing structure moment of inertia • Larger separation between facings for sandwich structure • From 4.7mm to 20mm • This complicates cooling and requires internal structure to provide high core shear modulus • Entirely different structural approach? • Any approach where support is provided at two opposing ends becomes a beam problem • Preferred solution is to provide a mid-span support at 1m • Allows retention of basic construction concept VG 23

24. Closing Remarks on Stave Distortion • Sag from lack of stave sandwich stiffness is not the problem! • Possible devise a pin-end support that controls sag---defined by near fixed end support • First sag problem only for upper and lower staves • Secondly, 1m long stave mass fully loaded is of the order of 5N (~1lb) • Reactions at pins very small, distortion of end caps not the issue • What are the problems? • Distortion or lack of dimensional accuracy in the end support plate surface • Any slight imposed “moment” on the stave end will produce large out-of-plane distortion • Angle at the stave end of just 0.04º will cause 250 microns at the center • This angle is equivalent to what occurs with a simply supported beam • A necessary step is an adjustable mount for the end cap • Stability of the end plate support---will become an issue VG 24

25. Closing Stave Remarks (continued) • End Plate Support • Factors to be considered in designing stave end support • Structure CTE • If positive or negative, how much? • For 60º temperature change, 1ppm/ºC will cause 60microns in a 1m length • Will a change in support structure length distort the stave mounting surface? • Structure CME • A quasi-isotropic composite (75Msi fiber modulus) with a cyanate ester resin will have a CME of 20ppm at 55% RH, or 20microns in 1m • Conceivable combined strain will be ~80 microns or more • Temperature differential across the stave • Need to determine out-of-plane movement for temperature gradient normal to stave axis • May not be easy to identify magnitude of the gradient VG 25

26. Thermal Solutions

27. Thermal Property Parameters FEA Model Properties Suspect the cable bus thermal conductivity is somewhat low; thermal property will be determined in order to improve thermal quality of results VG 27

28. Revised 4.7mm Flat Tube Model (No Wafer Heat) Cooling tube wall -22ºC • Separation between facings 4.7mm- present prototype • Modified hybrid stack, now---- • Chip, followed by silver adhesive, sitting on dielectric (5W/mK), followed by silver adhesive, sitting on BeO (240W/mK) • Before the dielectric and BeO were merged into one solid (8W/mK) • Cable K=0.4W/mK to simulate 75µm of Kapton, whereas model thickness is 250 microns • Coolant (-25ºC, 1800W/m2K) • Film coefficient to produce about 3ºC temperature drop • Results, average chips -15.5ºC; module center area -17ºC to -17.8ºC Chips (average) Detector center area Solution with CFDesign- CFD code VG 28

29. Revised 4.7mm Flat Tube Model (Add 4W per Wafer) • Separation between facings 4.7mm • Modified hybrid stack, now---- • Chip, followed by silver adhesive, sitting on dielectric (5W/mK), followed by silver adhesive, sitting on BeO (240W/mK) • Before the dielectric and BeO were merged into one solid (8W/mK) • Cable K=0.4W/mK to simulate 75µm of Kapton, whereas model thickness is 250 microns • Coolant (-25ºC, 1800W/m2K) • Film coefficient to produce about 3ºC temperature drop • Results, average chips -13ºC, module center area -14.5ºC to -15.3ºC Chips average Detector center area Solution with CFDesign- CFD code VG 29

30. Thermal Solution with NASTRAN -25ºC Coolant • Latest stave model- section one silicon detector in length • Double-sided detector • 0.5W per chip • No interface resistances Ends affected by cut-off Chip ~ -16ºC H=1800W/m2K Same parameters as in previous slide Silicon module ~ -17.8ºC Simple solution to recover strains VG 30

31. Gradient Through Stave Materials • Could improve contact with tube, but resistance is in the cable • Solution with CFDesign includes the contact interface resistances chip dielectric hybrid BeO cable facing module Extracted from NASTRAN FEA-chip heat only VG 31

32. Carbon-Carbon Laminate • For those that like C-C laminates as opposed to resin matrix laminate • Changing the through the thickness K from 1.44W/mK to 10W/mK we get approximately a 2ºC change in chip peak temperature (-15.9ºC versus -17.9ºC Thermal block is in the cable! Chip heat only VG 32

33. NASTRAN Thermal Strains From -25ºC Coolant • Latest stave model- section one silicon module in length • Double-sided module • 0.5W per chip (temp solution from slide #30) Peak out-of-plane distortion ~5microns VG 33

34. NASTRAN Thermal Strains From -35ºC Coolant • Latest stave model- section one silicon module in length • Double-sided module • Estimation of tube strain for 60ºC temperature change---- 6.2microns VG 34

35. Stave Thermal Strains- Old Model • FEA solution- alternating modules • Used 48cm long stave, symmetry BC at mid-length • Limit to stave length permitted addition of wafers and hybrids, still model required large number of elements • Thermal strain for change 50ºC in temperature (RT to -25ºC • Results for stave with 5.88mm core height, round tube • K13D2U 4:1 fiber orientation • 0.75mm thick facings • Detector modules and hybrids coupled to facing without interface compliance • Distortion pattern • ~11micron peak to peak, alternating in z-direction Balanced double-sided module stave strain is about half VG 35

36. Stave Thermal Solution – Old Model • FEA solution for both Al and Carbon-Fiber Filled PEEK tubes • 6 chips for each hybrid, 0.5W each: heat zone shifts top to bottom • Film coefficient 3000W/m2K, typical of Pixel stave for C3F8, two parallel tubes • Lower peak wafer temperature favors Al cooling tube (Unfilled PEEK) Silicon -11C Chips -7.1C -17.5C Chip/Silicon -23.5C -13.4C Chip/Silicon -20.2C VG 36

37. Evaporative Film Coefficient • Thermal FEA based on film coefficient from ATLAS (3000W/m2K) • 3000W/m2K chosen to expedite the earlier thermal solutions • Question: what might the film coefficient be for various diameter tubes? • Using Kreith-Bohn • As a check, calculated for 3mm (8mil wall) round tube (ATLAS) Dh=2.59mm • 3040W/m2K at entrance (x=0.3 after throttling) • 4451W/m2K at exit (x=0.8, mostly vapor) • Based on 4.6mm flatten tube (FEA thermal solution) Dh=4.956mm • 1228W/m2K at entrance and 1645W/m2K at exit, noticeably lower • Now for the 7.33mm flatten tube --Dh=7.791mm • 811W/m2K entrance and 1000W/m2K at exit • Important to note that since ΔTfilm is proportional to inner wall surface area, the lower heat transfer film coefficients have no real impact for same heat load • 2.42ºC (3mm), 2.96ºC (4.6mm), and 2.94ºC (7.33mm) • The pressure drop is different, lower for the bigger tubes, since the mass flow rate is the same for each case VG 37

38. Coolant Pressure Drop and Convective Heat Transport Objective: 1st order calculations for the semi-flatten tube and circular tube for both C3F8 and CO2

39. Tube Hydraulic and Thermal Evaluation • Stave with C3F8 Coolant: Semi-Flattened tube • Assumptions: entering quality 5%, exiting 75% • Resulting mass flow for 192W, 2.7g/s • 4.6mm semi-flattened tube, Dh=5.0mm • Entering Fluid -25ºC and 1.67bar • Average frictional pressure drop 0.137bar, resulting saturation exiting fluid temperature becomes -27ºC • Average film coefficient is 1800W/m2K • Conclusion • From hydraulic and heat transport standpoint the tube is slightly too large • Had to update the thermal solution for lower than estimated film coefficient • Original estimate based on higher coolant inlet quality and higher mass flow rate (injection for ATLAS seemed to be based on 30% quality) VG 39

40. Tube Hydraulic and Thermal Evaluation • Stave with C3F8 Coolant: Round tube • Assumptions: entering quality 5%, exiting 75% • Resulting mass flow for 192W, 2.7g/s • 4.6mm round tube, Dh=4mm • Entering Fluid -25ºC and 1.67bar • Average frictional pressure drop 0.41bar, resulting saturation exiting fluid temperature becomes -32ºC • Average film coefficient is 2500W/m2K • Conclusion • Coolant temperature varying between -25ºC and -32ºC helps to hold silicon temperature • Similar to having average coolant temperature of approximately -28.5ºC • From hydraulic standpoint the tube is slightly small • Average predicted frictional pressure drop encroaches on requisite vapor return pressure VG 40

41. Tube Hydraulic and Thermal Evaluation • Stave with CO2Coolant: Round tube • Assumptions: entering quality 5%, exiting 75% • Resulting mass flow for 192W, 0.82g/s • 4.6mm round tube, Dh=4mm (satisfies our core height issue in a sandwich to minimize gravity sag) • Entering Fluid -35ºC and 12bar • Average frictional pressure drop is negligible, 0.034bar, essentially no change in saturation temperature over the 2m tube length • Average film coefficient is 4650W/m2K • Conclusion • From hydraulic and heat transport standpoint the tube is more than adequate • Stress, for a 1000 psi max pressure is 6545 psi, which would be considered allowable for 3003 work-hardened Aluminum tubing (or age-harden Al alloy) VG 41

42. Hydraulic and Thermal Summary • Silicon Module Temperature for 2m (round trip) stave with 192W heat load • If module requirement is -25ºC, the coolant most likely must enter at -35ºC or at least minus 30ºC. • C3F8 saturation pressure at -35ºC is 1.04bar, which means a 0.67bar reduction in available return pressure to the compressor from ATLAS conditions, may be a system problem • Is -30ºC acceptable from system standpoint? • Frictional pressure loss in a 4.6mm semi-flatten tube with a hydraulic diameter of 5mm at -35ºC coolant inlet is estimated to be 0.18bar, further contributing to return pressure problem • For a “hard” silicon temperature of -25ºC, CO2entering at -35ºC is preferable • Prospects for C3F8 at -25ºC---most of problem comes from the 192W heat load • Need to improve thermal modeling by quantifying the thermal conductivity of the components • Consider strategy of augmenting heat spreading beneath the chips VG 42

43. Coolant Tube Stress AnalysisCompare Semi-Flatten Tube versus Round Tubes

44. High Pressure Coolant Tube • Geometry of coolant circuit modified to accommodate a round tube • Aluminum tube: 4.6mm OD (4mm ID), hoop stress for 60bar (1000psi) would be 6546psi (3003 Al has adequate strength) • Saturation of CO2 at 25ºC is 933psi. Foam pieces Round tube is supplemented with graphite foam transition to enhance thermal contact VG 44

45. FEA-Section of Tube • Tube Shape • Both prototype stave tube and tube for the “proposed stave” geometry have a semi-flatten shape • Flat portion is ~2mm across, to enhance thermal contact. • Both tubes are Al tubes with 12mil wall • For simplicity the shape of the sides are assumed to be circular with radius defined by core height (distance between composite facings) VG 45

46. 4.6mm Semi-Flattened Tube • NASTRAN solution • Internal pressure of 100 psi • Stress bar is Pascal; average von Mises stress in middle between two truncated faces is 6062 psi. • Deflection of corresponding node in this region is 15.5 microns. • 100 psi pressure is equivalent to 6.9bar • Pressure of C3F8 at -25ºC is 1.67 bar and at 25ºC the pressure is 8.7bar • Allowable yield stress for work-hardened 3003 Series Al is ~27,000 psi • But what happens in the stave? Middle is region of interest Stress at ends affected by end conditions VG 46

47. Prototype Stave -4.67mm Height • Present Design----Double-sided module • Partially flattened tube, 4.67mm height • 100psi internal pressure load • Peak deflection of stave surface ~1.5microns, versus nominally 15.5microns for unsupported tube (previous slide) • Results in much lower tube stress NASTRAN FEA VG 47

48. Stave Prototypes

49. Stave Prototype – Slightly Flatten Tube • Flat portion • 2mm flat to improve composite facing thermal contact with tube • Thermal adhesive • CGL7018, to reduce axial thermal strain from CTE mismatch • Option: Use EG7658 semi-rigid “filled” adhesive with grafoil to accommodate thermal mismatch VG 49

50. Stave Prototypes-Round Tube • Stave with round tube for high pressure CO2 • Heat transfer from facing to tube enhanced with graphitic foam • Typical values quoted: 150W/mK through thickness, 50W/mK in-plane • Free space occupied by graphite fiber honeycomb • Stave length and separation between facings in both prototypes Foam pieces Separation between facings same in both staves VG 50