1. M. Gilchriese Integrated Stave Mechanics and Cooling ATLAS Upgrade Workshop December 2007 M. Cepeda, S. Dardin, M. Gilchriese, C. Haber and R. Post LBNL W.Miller and W. Miller iTi

2. M. Gilchriese 2 Nominal Design – Short Strips Bus cable Hybrids Coolant tube structure Carbon honeycomb or foam Carbon fiber facing Readout IC’s 10cm detectors, each with 4 rows of chips on hybrids Hybrids glued to detectors and this assembly glued to mechanical/cooling support ~ 1 meter

3. M. Gilchriese 3 What Has Been Done Prototypes fabricated and studied –1m long for mounting silicon modules(see talk by Haber) –Three thermal prototypes, each about 1/3m long, with heaters, silicon, cables, dummy hybrids to simulate nominal design Three different tube types(to simulate compatibility with C3F8, C2F6/C3F8 mixtures and CO2) Varied facing thickness and adhesives Thermal measurements(IR imaging) before and after thermal cycling from 20C -35C fifty times completed Measured weights as input to material calculations –Preliminary demonstration of removal/replacement of silicon on stave completed. Design studies and FEA –Primarily of nominal design(hybrids glued top of silicon, short strips) Thermal performance and thermal runaway Gravitational deflections and support concepts Thermal distortion (as detector is cooled down) Long strips ie. outer barrels. –But also some work on Bridged hybrid Other – hermeticity, endcaps, production, R&D plan, risks……….. Summary only here – see Backup slides and references therein

4. M. Gilchriese 4 Prototypes We chose to build three prototypes to approximately simulate tube sizes that would be applicable to C3F8, C2F6/C3F8 and CO2. To answer the question - is this design compatible with all these? Tube diameters chosen from practical concerns(ie we had tube or could easily get it) but also from calculations of fluid properties that are summarized in the table below.

5. M. Gilchriese 5 Prototype Construction 1m prototype 2.8mm tube/foam 4.9 mm tube/foam Flattened tube Note prototype width is about 7cm – set in 2006 POCO foam: about 0.5 g/cc thermally conducting carbon foam Facings are K13D2U fiber laminates Carbon honeycomb All tubes aluminum

6. M. Gilchriese 6 Thermal Prototypes Water at about 20C IR images Before and after thermal cycling between 20C and -35C 50 times Bus cable Alumina hybrids Heaters 0.3mm silicon 3.3 W/hybrid(0.55 W/chip) No Power

7. M. Gilchriese 7 Major Prototype Lessons Thermal performance -  T between coolant and dummy detector –Same after thermal cycling to -35C fifty times. No evidence of lost coupling of tube to facing. –Same within measurement error for facing thickness in range about 0.25 – 0.7 mm. This also validated by FEA calculation. –Same within about 15% for all three tube types, also expected from FEA. Better for round tube with foam. –  T measured agrees with  T FEA to within about 1.5C or better, so can have some confidence in FEA Deflection measurements (of 1m prototype) agree with calculations within about 15%. Multiple successful trials of gluing dummy silicon to bus cable with SE4445(thermally conducting, flexible adhesive used to attach current pixel modules), removal using simple tooling(essentially a guided wire), clean up and reattach at same spot. See Backup for the details

8. M. Gilchriese 8 Thermal Model – 10cm Detectors Nominal design, hybrids glued to top of silicon Section shown below. Detailed properties in Backup Detector heating from parameterization of Unno at MIWG 11/07 Vary chip power(0.125-0.5W/chip) and detector power 1 or 2 mW/mm 2 at 0C Vary wall temperature of cooling tube to simulate different coolants and to see runaway.

9. M. Gilchriese 9 Thermal Performance C3F8 CO2 C2F6/C3F8 Note in this design, chip temperatures are within <2C of detector temperature

10. M. Gilchriese 10 More Thermal Performance Variants to lower temperature(with either more material or more cost or both) –“Triple-U” tube(see picture below). This would decrease the  T by very approximately 4C in the temperature region of interest for the nominal design. –Carbon-carbon(that has much higher thermal conductivity through thickness) but gain <2C and not worth much higher cost. –Small gain(~1C) possible with fiber(eg. K1100 instead of K13D2U) that has higher axial thermal conductivity(by about 40%) but somewhat higher cost. –Thermally conducting foam rather than honeycomb would gain 2-3C. –Multi-part bus cable(not cover all of detector area). TBD Conclusion from these studies depends obviously on coolant, chip power and prediction of detector power –If CO2, can have significant headroom –If C3F8, headroom reduced compared to CO2(no surprise) by ≥10C but still may be acceptable. –If C2F6/C3F8 mixture somewhere in between In all cases, need to validate with measurements on full-scale prototypes - - - Note that  T is referenced to tube wall temperature in this talk

11. M. Gilchriese 11 Thermal Model – Bridged Hybrid Wire bonds, simulated as thin solid, reduced K to 97/mK Chips 0.38mm thick (148W/mK) Al Cooling tube 0.21mm ID Separation between facings 4.95mm 10cm Foam bridge support 1mm air gap for bridge

12. M. Gilchriese 12 Bridge – No Air, No Wire Bonds -28C tube wall, 0.25W/chip and no detector power Maximum detector temperature about -20C But chips much warmer And gradient of chip T from center of bridge to edge is about 10C Next try to add wire bonds and air conduction(not trivial)

13. M. Gilchriese 13 More Bridged Hybrid Enclosed model in “box” to estimate effects of air conduction and include wire bond conduction. Preliminary result for -28C wall temperature, 0.25W/chip and zero detector power is max detector temperature -23C to -26C. Depends on model technique – see Backup. Chip temperature only about 2C higher than detector T Air/wire bond conduction important, studies continuing to validate Air box

14. M. Gilchriese 14 Material The total radiation length of the mechanical/cooling structure depends on the support conditions. If one assumes a shell-like support, then just the stave mechanical/cooling structure (for the nominal design) has a radiation length of about 0.4% (based on measured weights of the prototypes). This does not include coolant, modules or bus cables and most importantly not the shell or supports to it. For reference, if one assumes end-supports only (not shells) 1m meter apart, then the stave radiation length for comparable deflection would be about 0.9% because stave has to be stiffer. In short, one needs a combined stave-support-structure design to get a realistic number.

15. M. Gilchriese 15 Miscellany Outer barrel, long strips, with single hybrid for 10cm detectors. –Quick look at thermal performance and  T to hottest point on silicon is about 3C for 0.25W/chip and 0 detector power in nominal design. More than expected from ¼ scaling by power but <1/2 of short strips. Distortions as detector cools down –<10 microns distortion out-of-plane calculated for about 50C temperature change but only done so far for 0.7mm facings. Need to do for thinner facings. Gravity sag –A non-issue with shell-like support and frequent supports. Tune structures to meet requirement.

16. M. Gilchriese 16 Hermiticity and Endcaps Hermiticity –In nominal design, there are gaps(eg. 100 microns) between adjacent detectors on same side of stave. –Mechanically at least one could offset(in Z) the top and bottom detectors such that overlaps do not line up. –But what can you do if you want hermiticity on each side by overlapping detectors in Z? –Simplest mechanical solution is to have ½ the modules glued to bus cable and the other half with about 0.5mm carbon fiber plate(with appropriate cutouts) glued to bus cable, raising ½ of the modules to allow overlap – see cartoon below.  T would increase for these <1C (for 0.25W/chip, 1mW/mm 2 nominal design, not bridge) Endcaps –Purely from a mechanical/cooling aspect, one could make pie-shaped segments, routing the cooling tube as needed to achieve the same thermal performance as barrel staves. Likely offset slightly front and back facings to avoid complete gaps. Straightforward (as would be the support of these into a disk) –But, but this doesn’t address the considerable problems associate with detector layout or bus-cable-like connections….. Z

17. M. Gilchriese 17 Interface to Barrel Support Have looked quickly at different support options Current preference is for shell-like overall support with support points about every 50 cm. More on this topic in the engineering session. Note that this implies strong coupling of stave design with shell eg. where is the stiffness and not just the obvious support interfaces. Light weight composite sandwich rings Locating pins in rings

18. M. Gilchriese 18 Production – Mechanical/Cooling ONLY Rough numbers for barrel only(including guess for yields) –250 1m staves –300 2m staves Materials for nominal design –100 kg of fiber unitape –100 m 2 of honeycomb or structural foam –10 5 cm 3 of thermally conducting foam (if used) –600 tube sets(tubes, fittings) Production rate 1-2/day required At minimum, industrial production of subcomponents (facings, honeycomb/foam pieces, bare tubes…) but full industrial production feasible (with substantial lab QC oversight). Very rough guess for technical production labor(not including design or prototype phase or cable attach or anything related to module attachment) is 15 FTE-years

19. M. Gilchriese 19 Risks – Mechanical/Cooling ONLY –Identify components which are ‘single-source NONE for nominal design. –List components or assembly steps which need to be developed or which need qualification before they can be adopted e.g. Critical decisions before starting real design: what is coolant and bridged hybrid or not Interfaces(primarily electrical, where are the connections to electrical services) –How deep is the assembly pipeline? i.e. what level of prototyping effort is required to fully validate design? Would clearly have to build full-scale prototypes for thermal performance/stability –Risk due to tight mechanical tolerances during assembly Moderate, mitigate with QC (CMM) –Grounding & shielding risks (by external experts) e.g. See electrical department…… –Options available for repair i.e. the risk of damage being unrepairable. Module/stave damaged during shipping or testing Replace module or toss out Module/stave damaged during assembly to structures Replace module or toss out Cooling channel damaged during or after assembly. During, replace, after toss out(before module attach) Service cable damaged during/after assembly Don’t know yet Failure after integration. If design requirement(I don’t think it is), make stave replaceable –Flexibility or options to modify the design based on experience with ATLAS. Could the material be reduced further? And at what cost? Some but more design time and prototypes Could additional measurement layers be introduced? Sure Are there options to improve the mechanical stability? Easy but more material Are there options to improve the cooling i.e. remove more power and/or lower temperature? Yes, many Are there options to improve grounding and shielding? Talk to electrical department –Other technical, schedule or financial risks Need more studies of pipe(Al) in potential contact with carbon, although lots of work already done for existing pixel system Mechanical/cooling ONLY part is of moderate risk. Not easy but not hard

20. M. Gilchriese 20 R&D Plan – Mechanics/Cooling ONLY If this approach, then what would be development plan? Need to design, build and test realistic prototypes(eg. 1m length) that simulate well thermal and mechanical expectations. But it does not make sense to do this before making some essential decisions –Coolant –Hybrid type(bridged or not) –Hermeticity –At least preliminary definition of critical interfaces(mounts and electro-optical) Given these choices, estimate it would take about one year to design, build and test full-scale prototype(s) for thermal and structural performance Concurrently could develop preliminary production plan and associated cost

21. M. Gilchriese 21 Next Steps Respond to questions from review panel Apart from this does not make sense to do much more mechanical/cooling design work or prototype fabrication for this concept in the next months. Need feedback from making 1m functional prototype with real modules And decisions on requirements and constraints

22. M. Gilchriese Integrated Stave Mechanics/Cooling Backup ATLAS Upgrade Workshop Valencia December 2007 M. Cepeda, S. Dardin, M. Gilchriese, C. Haber and R. Post LBNL W.Miller and W. Miller iTi

23. M. Gilchriese 23 Introduction We collect here some backup information for the presentation on integrated stave mechanics/cooling. A few notes –Work on the integrated stave began in the Fall of 2006 –The dimensions of prototypes, and a number of FEA calculations, were set then when detectors were assumed to be about 6cm in width. –Thus prototypes were built assuming about 6 cm wide detector dimensions rather than the current 10cm “baseline”. Thus a principal goal of the “6 cm” prototypes is to validate FEA estimates of the thermal performance, and then use the FEA to calculate for 10 cm –In addition, the properties assumed for materials, particularly for thermal FEA calculations have evolved somewhat with time as have assumptions for detector power after irradiation. Link to information on integrated stave mechanics/cooling http://phyweb.lbl.gov/atlaswiki/index.php?title=ATLAS_Upgrade_RandD_-_Mechanical_Studies

24. M. Gilchriese Prototypes

25. M. Gilchriese 25 Reminder of Prototype Concept 71.5mm For prototypes……..fixed > 1 year ago K13D2U, high-modulus facings Adjust facing thickness(layers) to achieve stiffness desired Carbon-fiber honeycomb in-between facing, fixed thickness Three types of tubes –Flattened(C3F8) –Big round with POCO foam(C3F8/C2F6) –Small round with POCO foam(CO2) POCO foam: about 0.5 g/cc thermally conducting carbon foam Link to drawings is here

26. M. Gilchriese 26 Prototype Stave Core Assembly Length (m) Facing Material # of Plys Facing Tube Type PurposeStatus 10.35CN6010FlattenedAssembly trial Complete 20.35K13D2U10FlattenedShort, thermal prototype Complete 31.0K13D2U10FlattenedFor modules Complete 40.35K13D2U34.8 mm round/ POCO foam Foam bonding, thermal prototype Complete 50.35K13D2U 32.8 mm round/ POCO foam CO2 thermal prototype Complete 6?K13D2U???TBD in 2008

27. M. Gilchriese 27 Weight and Material Measured weights for 1m prototype(10 ply facings) and extrapolation to thinner facings(3 ply) and width for 10cm detectors given below. Note assumes minimal side closeouts Tube is flattened. Would get similar numbers for POCO foam+smaller tube

28. M. Gilchriese 28 Thermal Measurements Measurements before and after thermal cycle 50 times to -35C are summarized below –Delta T calculated from average of inlet+outlet water T for convenience. Max and min given to nearest 0.5C. Delta T rounded to nearest degree. –No difference between before and after thermal cycle within errors –Note tube(4.8) with foam compared to flattened is better as is smaller tube with foam. We attribute this to better coupling to tube FEA results are given(for fixed fluid temperature everywhere). Agreement within 20% or roughly 1.5C. Writeup of FEA is at link herehere

29. M. Gilchriese 29 Remove/Replace We have completed a number of trials of gluing glass and silicon with SE4445 adhesive that was used to attach all pixel modules to local supports in the current pixel detector. Has decent thermal properties and already tested to 50 MRad for pixels.SE4445 Attach, let cure(both week long and about 2 month long tested), remove, clean and replace. Straightforward mechanically, only need simple tooling for close-together detectors – promising (no surprise since did this already for pixels) Pictures on next pages, although hard to see

30. M. Gilchriese 30 Removal Pictures Glass slide after removal(slide at bottom of picture) Starting to peel SE4445 Silicon detector after removal and before cleanup After about 2 month cure. Done with two detectors, same result

31. M. Gilchriese Thermal FEA

32. M. Gilchriese 32 Comments Some of the most recent results are included here Many previous studies with somewhat different parameters. See the wiki http://phyweb.lbl.gov/atlaswiki/index.php?title=ATLAS_Upgrade_RandD_-_Mechanical_Studies

33. M. Gilchriese 33 Thermal Runaway in 10cm Module Thermal Runaway Issue: Based on new detector heating curve- (revised by Nobu-MIWG meeting November 2007) –Quarter section from 10cm wide stave, single U-Tube –Spacing of U-Tube divides heat load collected by each symmetrically –Chip heat load and surface heating treated as variables

34. M. Gilchriese 34 Thermal Runaway Model Parameters

35. M. Gilchriese 35 Surface Heating Curve New curve based on 1mW/mm 2 at 0ºC (Nobu-MIWG Nov. 2007) and exponential temperature dependence

36. M. Gilchriese 36 Thermal Runaway Solutions Plot of peak detector temperature leading up to runaway (as function of tube surface wall temperature) Surface heating 1mW/mm 2 @ 0C Exponential temperature dependency (Nobu-MIWG Mtg. Nov. 2007)

37. M. Gilchriese 37 Thermal Runaway-Variable Surface Heating Comparing effect of surface heating using 0.25W/chip as baseline Surface Heating 0 1mW/mm 2 2mW/mm 2

38. M. Gilchriese 38 Detector Surface Heating Curve at right shows slight deviation of solution convergence Deviation caused by using peak silicon nodal temperature whereas solution is based on the detector outer surface edge average

39. M. Gilchriese 39 Thermal Runaway-Typical Thermal Plot Chip: 0.5W Coolant Tube Surface -16.8ºC Peak chip: 6.18ºC Peak detector edge: 5.17ºC Throughout solutions peak chip and peak detector differential temperature stays near 1.0 to 1.1ºC With 0.25W/chip the temp difference is nominally 0.5ºC Nearly thermal runaway point

40. M. Gilchriese 40 Bridge Thermal Model Salient Features –High conductivity (700W/mK, 0.5mm thick) CC bridge material support for 0.28mm thick hybrid(1W/mK) –40 chips @ 0.25W/chip –Detector 0.28mm thick, 148W/mK –Allcomp carbon foam for bridge support (isotropic 45W/mK) –Carbon Foam for tube support (45/45/45 W/mK) Reduced density over POCO foam (0.2g/cc versus 0.5 g/cc) –Sandwich foam Allcomp foam option, ~0.1g/cc @ 3W/mK Comparison with Hybrid on 10cm Detector –Thermal solution with both with inner tube wall at -28 º C Simulates -30 º C with 8000W/m 2 K No change made to material properties in 10cm detector with integrated hybrid

41. M. Gilchriese 41 10cm Detector-No Bridge Material Properties –See previous slide (#2) 40 chips per detector, 80 total –0.25W/chip Q (Si)=0W –Tube inner surface -28 º C, no convection coefficient Interest in ΔT from chip and detector surface to tube surface Peak chip temperature –Middle hybrid region: -20.5 º C Peak Detector –Middle hybrid region: -21.5 º C –ΔT in region of max gradient: 6.5 º C

42. M. Gilchriese 42 10 CM Wide Stave-No Bridge Solution –Replaced honeycomb core with Allcomp carbon foam (<0.2g/cm 3 : 45W/mK) –Also, replaced POCO foam tube support with same foam Peak Chip Temp: -22.7 º C Peak Detector: -24 º C –ΔT (referenced to tube wall) 4 º C

43. M. Gilchriese 43 10 CM Wide Stave-No Bridge Solution: Simulate “outer” long strip detector –One upper and power hybrid for 10cm detector –20 chips @ 0.25W/chip –Coolant tube inner surface: -28 º C –Materials, see slide (#2) Detector –Peak temp beneath hybrid: - 24.8 º C –ΔT in region of max gradient: 3.2 º C Chip Peak Temp: -24.1 º C

44. M. Gilchriese 44 Thermal Bridge Model (1/2 of 10cm) Wire bonds, simulated as thin solid, reduced K to 97W/mK Chips 0.38mm thick (148W/mK) Al Cooling tube 0.21mm ID Separation between facings 4.95mm 10cm Foam bridge support 1mm air gap for bridge

45. M. Gilchriese 45 Bridge Thermal Model Enclosed bridge model in an air box. Air participates only through pure conduction. Air fills all cavities not occupied by a solid Air box

46. M. Gilchriese 46 Model Parameters Cable and adjacent adhesive layers modeled as single layer 0.227mm and K=0.31W/mK

47. M. Gilchriese 47 Solution with -28C tube wall 0.5W/chip Q (Si)=0 Slight asymmetry caused by variance in interior coolant wall temperature Detector max=-21.4ºC Chip peak=-16.5ºC

48. M. Gilchriese 48 Solution with -28C tube wall 0.25W/chip Q (Si)=0 Slight asymmetry caused by variance in interior coolant wall temperature Detector max=-25.8ºC Chip peak=-23.3ºC

49. M. Gilchriese 49 Solution with -28C tube wall 0.25W/chip Q (Si)=0 Bridge foam and tube foam 45W/mk, density ~0.2 g/cm3 (no POCO foam) Peak detector temp -24.2ºC Sandwich foam core 3W/mK, density ~0.06 g/cm 3 Peak chip=-21.8ºC Wire bonds 97W/mK

50. M. Gilchriese 50 Bridge: Complete Model Two orthogonal cutting planes through air box

51. M. Gilchriese 51 Bridge: Complete Model Transverse cutting plane through air

52. M. Gilchriese 52 Bridge: No Air-No Wire Bonds Shaded with Mesh Shaded Without Mesh Gradient from peak chip to edge of bridge=10.2C -28C tube wall, 0.25W/chip, no detector power

53. M. Gilchriese 53 Bridge Model-1/4 Size Solution without air box and wire bonds –Previous model under same conditions gave 3.6C for peak chip temperature –See previous slide, this model resulted in 4.5C Solution considered close enough for running with effects of air and wire bonds and faster than larger model

54. M. Gilchriese 54 Bridge Model-1/4 Size with Air and Wire Bonds Quarter model 895560 fluid elements 584136 solid elements 96069 adhesive elements Thermal solution With air and wire bonds Similar results to half size model Detector peak temperature -23.1C

55. M. Gilchriese 55 Fluid Calculations C3F8 calculations are here for flattened tube and here for round tubeflattened tuberound tube CO2 calculations are here and here.here Summary from main talk reproduced below Note  T(film) is an average around the loop  T(loop) follows from the P vs T curves for the fluids and is rounded to the nearest 0.5C These calculations are complex and need validation by measurements

56. M. Gilchriese 56 Adhesive Joint Considerations There are numerous analytic solutions for adhesive joint shear stress caused by thermal expansion of dissimilar materials –General theme is that the shear stress is a maximum at the ends of joint, and essentially zero at the center –Maximum shear stress at the end is independent of the length of the joint Key factors are: – modulus of elasticity, CTE, and thickness of joined materials –thickness and shear modulus of the adhesive –Temperature differential A useful reference to bound the problem: Thermal Stresses in Bonded Joints, W.T. Chen and C.W. Nelson –Suggests for carbon foam joined to aluminum tube with CGL7018 (very compliant adhesive) or EG7658 (semi-rigid) that shear stresses remain within material limits for a 100C temperature change –Prototype testing will confirm our expectations

57. M. Gilchriese 57 Carbon Foam to Aluminum Tube Joint 100C temperature differential –Cure temp to -25C –Foam thickness=8mm, G=690MPa, α=4ppm/C –Aluminum wall thickness 0.305mm, E=10Msi, α=12ppmC –Adhesive thickness=0.10mm, Compliant G=40MPa (5862psi), Rigid G=1 GPa Max shear stress, τ=1062psi, compliant τ= 42psi

58. M. Gilchriese 58 Computer-Based Solutions Structural Problems –NASTRAN FE solver Recent solutions with NE NASTRAN with FEMAP interface Prior work with MSC NASTRAN, but MSC no longer can bundle the NASTRAN solver with FEMAP pre-processor –Choose not to use PATRAN pre-processor Fluid/Thermal Problems –Use CFDesign computational fluids dynamics code Very versatile Allows use of shell elements for describing interface resistances HEP Silicon-Based Tracking Detectors –Issue with very, very thin solids mixed in with larger solids In reasonable sized geometry, some solids may have only surface nodes, and no internal nodes; –possible consequence is reduction of solution accuracy