Understanding the deformation issue of the ATLAS IBL

Understanding the deformation issue of the ATLAS IBL detector F. Cadoux, S. Coelli, D. Giugni, C. Hsu, S. Michal, H. Oide, E. Vigeolas Tracking Detector Mechanics 2015 1

Outline • • • IBL’s modules support cinematics Experimental evidence of the issue Theoretical model for the CTE mismatch effects on beam-like structures FEA studies Lessons learned Tracking Detector Mechanics 2015 2

IBL Lay-out • The stave supports the modules and provides thermal link to the cooling pipe in which the CO 2 boils off • Modules are facing the beam pipe • The layout counts 14 titled staves arranged in an “impeller-like” fashon (Length~700 mm) • The cross-section is triangular. Carbon foam is sandwiched between K 13 composite laminates Beam Pipe Carbon Foam CO 2 Cooling Pipe • The electrical services are Kaptonbased flex directly bonded onto the stave laminate. Service Flex Bus Pixel Module Stave Inner Support Tube Tracking Detector Mechanics 2015 3

IBL Global Support Pixel End ring ISIS IPT Beam Pipe Sliders Inner Support Tube R Reference • The IST (Inner Support Tube) is the IBL global support. • In the beam perpendicular plane IST is supported off the Pixel detector by means of the ISIS (Inner Support Interlink Structure). • The mechanical interface to the detector is provided by radial sliders Z and F Reference • IPT (Inner Positioning Tube) is mechanically grounded at PP 1 (~2. 5 m away in Z) and it sets the absolute Z and F location Tracking Detector Mechanics 2015 4

At Z=0 the stave is fixed in R but can rotate around Z On A-Side the stave end Can slide along Z C-Side Z=0 pin screw Stave Support Ring Spring Loaded screw Slotted Pin Support kinematic On C-Side the stave end is locked in Z Z axi s The IBL staves are therefore like a set of double constrained beams A-Side • Constrained in Z on the C-Side • Sliding along Z on the A-Side • Linked together in the middle by a ring Tracking Detector Mechanics 2015 5

Powering & cooling dynamic at e h e th f o 10% he flex t from • When the cooling is turned on, the CO 2 starts boiling and it takes all the stave down to the evaporation temperature. • This is the most demanding situation for the stave: The CTE mismatch between the materials has its effect maximized. CO 2 boiling Heat from the module The thermal gradient induced by the heat load mitigates the distortion reducing the T drop from the assembly temperature Note that this effect is marginal. T gradient is ~8 ºC when compared to the total cooling down range ~50ºC (from +20 C -30ºC) 6

Outline • IBL’s modules support cinematics • Experimental evidence of the issue • Theoretical model for the CTE mismatch effects on beam-like structures • FEA studies • Lessons learned Tracking Detector Mechanics 2015 7

Observed deformation ZL ZL XL XL XL Let’s draw the following coordinate system: LOCAL CARTESIAN • • XL axis laying on the sensor plane ZL axis perpendicular to the sensor plane GLOBAL CYLINDRICAL • R and F as in figure • Z along the beam axis ZL R F • Looking at the alignment for last cosmic rays data, relevant residual values along the XL axis have been found. • This leads to a displacement in the opposite direction that can be interpreted as a bow of the stave along the global F axis • The bow amplitude depends upon the evaporation set point. • The dependency is linear and the value is d=10. 6 ± 0. 7 mm/ºK 8

Outline • IBL’s modules support cinematics • Experimental evidence of the issue • Theoretical model for the CTE mismatch effects on beam-like structures • FEA studies • Lessons learned Tracking Detector Mechanics 2015 9

Some basics • In literature the problem is widely covered in the case of a beam loaded with a linear thermal gradient across its section • This is equivalent to a bi-metallic beam bonded at room temperature subjected to a uniform T drop T 1 Al Mf Mf T 2 sx T 1 T 0 Al Cu Al Mf Mf sx Al Cu • The stress status is equivalent and the deformation can be calculated for any constraint scenario Tracking Detector Mechanics 2015 10

Cantilevered beam with T gradient in the cross section • The deformation is: V. Displ [mm] • The apparent bending force: A quick simulation confirms it Tracking Detector Mechanics 2015 11

Double constrained beam with T gradient in the cross section That’s really puzzling… 1. Vertical reaction forces are all equal to 0 • We promptly thought that the bow were induced by CTE 2. If of the mismatch of different materials in the stave. In one particular by two ends can slide longitudinally the flex also the horizontal reactions are equal to 0 • While the titanium pipe is placed on the of symmetry of 3. axis. Internal bending force is constant along x and the stave section, the flex is off center and it induces an axial stress gradient, therefore a bending force equal to the bending reaction forces • Vertical Displacement [mm] ne g pla in Bend al Axi Flex s s xial Stre e. A Stav s s Stre BEAM DOES NOT BOW VERTICALLY! But analytical solution shows that a bending force cannot A quick generate the bow simulation confirms it! Tracking Detector Mechanics 2015 12

Where is the trick? • The solution stands on the fact the stave cross section is not constant along its length. At the ends peek inserts provide the mounting interfaces • The peek end-parts brake the “magic” of the double constrained beam subjected to CTE mismatch and it makes the stave bowing • Simple beam simulation shows how adding a short section of peek at the ends makes the bow to pop up again. V. Displ [mm] Peek sections 13

Understanding the kinematics • • We have seen that each stave tends to bow during the cool down due to the CTE mismatch (Flex w. r. t Stave) AND due to the peek interfaces at its ends The bending plane is nor parallel to the R-Z nor to the Z-F global cylindrical plane • However, all the stave are connected at Z=0 to a ring that can rotate around the Z-axis… therefore: e Fre ane ve ng Pl a t S ndi Be • • • The stave would bow in the “Stave Free Bending Plane” Since the arrangement is “impeller-like”, each one of the 14 staves contributes with an R and F components The Central Ring receives these forces: – The sum of the radial components over F is equal to zero. – While the F component sums up in a non zero twisting force Central Ring The central ring rotates along F (around the beam axis) All the detector sections (along Z) rotates rigidly with different amplitudes (generating the observed bow) and making the detector twisting symmetrically w. r. t. the global Z=0 plane 14

Outline • IBL’s modules support cinematics • Experimental evidence of the issue • Theoretical model for the CTE mismatch effects on beam-like structures • FEA studies • Lessons learned Tracking Detector Mechanics 2015 15

FEA simulations • • • The next attempt is running a FEA of the complete detector in order to match the stave deformations measured during the cosmic rays run. The problem here is related to the knowledge of the material properties of the stave’s parts. Most of them come from vendor’s datasheet and, mostly, they have not been directly measured Since the sag is heavily depending upon some parameters (like Flex CTE etc), the FEA ran until now could only give a qualitatively confirmation of the deformation modes. To match the simulated amplitudes to the measured values is a matter of the ongoing work… The estimate of the stress levels induced by the CTE mismatch to address the detector reliability at cold is the scope of the simulations. Unfortunately the predicted deformation does not match well enough the experimental data to address this issue yet. In particular, and preliminarily, there are two areas where there are indication that the stress level approaches or overcomes the adhesive yield: – It is the case of the Module to Stave interface – Stave laminate to peak Eo. S interface Tracking Detector Mechanics 2015 16

Preliminary FEA results • The detector twisting at the central ring appears evident in the simulations • On top of the F displacement at Z=0, the FEA has revealed also a radial bow. But this deformation mode cannot be appreciated with the tracks • The displacement of the sensor as function of Z follows the shape measured with tracks although the model requires some tuning to match the sag absolute value. Tracking Detector Mechanics 2015 17

Outline • IBL’s modules support cinematics • Experimental evidence of the issue • Theoretical model for the CTE mismatch effects on beam-like structures • FEA studies • Lessons learned Tracking Detector Mechanics 2015 18

Lessons learned (1) • • For the previous generation of ATLAS detectors, engineers tended to deal with the CTE mismatch problem making the parts floating each other. It works as long the interface is not a thermal interface. It is the case of the ATLAS Pixel detector where the cooling pipe is floating w. r. t. the stave structure. The thermal performance tends to degrade with the time and thermal cycles. For IBL the pipe has been bonded to the stave but the geometry must be such the resulting internal bending force is null or negligible. Mission was accomplished with the pipe but not with the flex In general services can be permanently bonded too (ATLAS ITk is evaluating co-curing flexes into the structures). but, again the flex with high CTE must be placed in the symmetry plane of the structure. When this does not happen, like for IBL where flex that has been bonded on the side of the stave, then the effects of the CTE mismatch can be difficult to predict and control. Tracking Detector Mechanics 2015 19

Lessons learned (2) • • Ganging mechanically several staves together at intermediate locations (by means of rotating rings, for example) relaxes the stiffness requirements on the local supports. But it also leads to coherent deformation modes (like the one discussed here) that can be appreciated only with global simulations (often hard to achieve) or complete mechanical mock-ups. In this case the effect was appreciated only when the first tracks have been produced. Fortunately thermal stability of the IBL cooling system mitigates the impact to the physics performance. Tracking Detector Mechanics 2015 20

Lessons learned (3) • • • Better engineering integration between the electrical services and the mechanical structure has been enforced during the last decade. Very often, in the past, the services have been treated as late add-on to the existing mechanical design. They usually come later (for many reasons) and they end up to be just adjusted. Now it’s no longer the case. The mechanic and electric souls are much more integrated in the design process. • However we see from the case described here that we must do better. The mechanical impacts of the electrical services should be evaluated at the earliest stage. At least earlier than what we thought was enough. • And let me conclude with a very last consideration we all knew already [Repetita Iuvant]: “Devil is in the details, indeed”… indeed” What described is just another spectacular proving case. Tracking Detector Mechanics 2015 21