1 MKU ATLAS CMS Alignment Tracker Alignment

1 MKU ATLAS & CMS Alignment Tracker Alignment Strategies for ATLAS and CMS Muge Karagoz Unel On behalf of ATLAS and CMS Alignment Groups 12 th April 2007 UK HEP Forum – LHC Startup Cosener's House, Abingdon

2 ATLAS & CMS Alignment Contents Will try to cover… • Motivations for alignment • The experiments and detectors • Alignment Performances • The current status and plans for early data Disclaimer: Will concentrate mostly on inner tracker alignment and make use of ATLAS. MKU Note: most trigger plans & early physics issues will be described by the other speakers.

3 ATLAS & CMS Alignment MKU the Prerequisites • General purpose LHC detectors ATLAS & CMS need to cope with demands of LHC physics programme requirements • Precision and accuracy is crucial for EWK and new physics • particle ID at ultra high energies • b-tagging for top and Higgs physics • W-mass measurement (one of the most challenging!) • Design parameters from ATLAS: • Examples from ATLAS tracking: • Calorimetry • local alignment < 10μm so as • (E) E 11. 4% E, for electrons • Tracking • (p. T/p. T) 20% , for muons of 500 Ge. V not to degrade intrinsic resolution > 20 % • B-field to 0. 1% locally • material globally to 1%

4 the ATLAS Detector ith 4 T toroid ometer (MDT) w Barrel m-spectr MKU ATLAS & CMS Alignment 22 m 46 m SC) r (MDT+C omete m-spectr oroid Endcap t with er (ID) Inner track on ) (TRT+silic d 2 T solenoi

5 the CMS Detector Barr el m (DT) MKU ATLAS & CMS Alignment 15 m Inn er 4 T track er es la ( so silic len o oid n) 22 m End cap m (C SC)

6 the Challenge: Atlas ID is BIG ATLAS & CMS Alignment TRT Pixels (3 layers+3 disks) 5. 4 m SCT endcaps: 9 disks SCT barrels: 4 layers Barrel TRT Forward TRT Sub. Detector Barrel Silicon Forward Silicon A/B/C PIX SCT # layers/disks/wheels 3 2 x(12/8/8) 3 4 2 x 3 2 x 9 # modules/planes 32 2 x 8 1456 2112 2 x 144 2 x 988 sub Total 96 448 MKU Total 544 • Silicon total Do. F = 6 x 5832 = 34992! • TRT total Do. F = 7 x 96 + 6 x 56= 1008 3568 2264 5832 Alignment challenge! 6 Do. F/module: 3 translations & 3 rotations

7 and CMS=Currently the Most Silicon Endcap (TEC) Outer Barrel (TOB) 9 discs, 4 -7 rings, 6400 modules ATLAS & CMS Alignment r (mm) 6 layers 5208 modules 206 m 2 of Si pixels not shown h blue = double-sided red = single-sided IP Inner Barrel (TIB) 4 layers, 2724 modules Inner Disc (TID) 3 discs, 3 rings, 816 modules MKU Resolutions: Strip pitch 80 -205 µm σ ≈ 23 -60 µm, 230 -520 15148 modules z (mm) µm Pixels size 100 x 150 µm σ ≈ 10 x 15 µm 1900 modules

8 ATLAS ID Module specifications ASIC s ATLAS & CMS Alignment MKU ~70 mm ~140 mm • Pixel detectors: real 2 -D readout – Size 50 400 m with 14 115(60) m resolution. • SCT modules: double-sided strip detectors with 1 -D binary RO/side (768 strips). – Strips pitch of 80 m giving 23 m resolution. – Stereo-angle of 40 mrad gives 580 m resolution in rz direction. – Mounting precision ~ 100 m – end-cap modules are in wedged shape • TRT has 300 k straw tubes – Size 4 mmx 740 mm, resolution 170 m (perp to wire)

9 Alignment Generalities ATLAS & CMS Alignment • Alignment is determination of – Sensitive detector position, orientation (6 parameters) – module deformation due to temperature, magnetic field, material load • For ex. , shrinkage of muon detectors of order of 1 cm with B-field on! • Consists of 4 components – Assembly knowledge: construction precision and surveys, for initial position corrections and errors – Online monitoring and alignment: lasers, cameras, before and during runs – Offline track-based alignment: using physics and cosmic data track residual information – Offline monitoring and alignment: using track and particle ID parameters • Challenges for the track-based alignment MKU – Both detectors have large number of Do. F to solve for. – Insensitivity to weak modes w/o additional constraints from data

10 ATLAS & CMS Alignment MKU Prospects for LHC Beams Parasitic collisions with wide range of interaction z-point With the recent magnet problems, not sure will happen in time or happen at all…

11 Expected Event Rates F. Gianotti (ICHEP 2006) Not much of anything else other than min bias and QCD jets • Large statistics of high pt muons within few weeks! Trigger studies are underway by both experiments ATLAS & CMS Alignment • Physics Run 2008 @ 14 Te. V, L~1032… 33 MKU •

12 Who Ordered Misalignment? Misalignment is due to Misalignment is time dependent! and when and how much the time the parts will move around is unknown. MKU ATLAS & CMS Alignment Precision of assembly Stress from magnetic field or thermal stress Changes due to humidity, … Martin Weber, CMS Misalignment studies: • Ideal geometry – No misalignment • Short-term (<1 fb-1) – First data taking – Hardware alignment used • Long term (1 -5 fb-1) – First alignment with highstatistics tracks, for first physics analysis • Final alignment – Do not deteriorate detector resolution

13 Misalignment: BSM Searches • Example for ~early LHC physics: Resonances in Di-Muons • 5 discovery reach for RS gravitons • Would need about 50% less data if optimal alignment! ATLAS & CMS Alignment C = 0. 01 (coupling constant) First data C =0. 1 Long term MKU Dimuon Mass Georg Steinbruck, CMS

14 Basic Tasks and Handles • MKU ATLAS & CMS Alignment • First days of data-taking: Figuring out anomalies: Calibration and alignment! First goal: working tracking reconstruction! – Hit errors, Dead/noisy hardware (and software!) components – Realistic simulation corrections, material effects – Match with muon chambers and calorimeters – Absolute momentum scale (using known resonances) – Tracking efficiency (dimuons from J/ , Upsilon, Z) CMS material corrections

15 Basic Tasks and Handles ATLAS & CMS Alignment • • • MKU • Alignment Handles: 2007 -2008 • Cosmic rays 2007 -2008 • Beam halo muons, beam gas events • Isolated muons from b decays, isolated 2007 -2008 tracks from MB events 2008+ • W, Z resonances • Note: Collision tracks and cosmics populate different parts of global covariance matrix of alignment -> make complete datasets Dedicated data streams Study timescales for detector movements and finalize the Software and Computing Model accordingly for long-term alignment Align first large structures, then sensors at high statistics or limit ourselves to limited number of Do. F Pre+during Pilot runs

16 Cosmics & Beam Halo MKU ATLAS & CMS Alignment Provided that they do not harm sensitive detector material! • ATLAS: Trigger using Tile. Cal, current trigger rates ~ 10 Hz • cosmics during commissioning, do not expect stable alignment until global cosmics (~fall 2007).

17 ATLAS ID Track-based Alignment • • ATLAS & CMS Alignment • • Intrinsic alignment of Silicon and TRT, Si+TRT, all rely on minimizing residuals Global 2: – minimization of 2 fit to track and alignment parameters – 6 Do. F, correlations managed, small number of iterations – Inherent challenge of large matrix handling and solving Local 2 : – similar to global 2 , but inversion of 6 x 6 matrix/module – 6 Do. F, no inter-module or MCS correlations – large number of iterations Robust Alignment: – weighted residuals, z & r overlap residuals of neighbouring modules – 2 -3 Do. F, many iterations, no minimization All algorithms implemented within ATLAS framework, sharing common tools Able to add constraints from physics & external data <- crucial! Tracks Digits MKU Iteration until convergence Reconstruction Alignment Algorithm Align. Constants Final Alignment Constants Performances on subsequent slides

18 ATLAS Global 2 Approach MKU ATLAS & CMS Alignment track Method consists of minimizing a giant 2 resulting from a simultaneous fit of all particle trajectories and alignment parameters: hit Intrinsic measurement error + MCS Use the linear expansion (assume all second order derivatives negligible). Track fit solved by: residual alignment parameters given by: Key relation! Equivalent to Millepede approach from V. Blobel for CMS

19 CMS Track-based ID Alignment • • Three different algorithms implemented in CMS reconstruction software Millepede-II: – Global 2 formalism – Replaces original Millepede (brute matrix inversion) with iterative solver – Most promising approach to CMS problem for long-term scenario MKU ATLAS & CMS Alignment • HIP Algorithm: – Local 2 , inversion of 6 x 6 matrix/module – correlations through iterations • Kalman Filter: – Iterative, based on Kalman filter update – Converges slower • Similar to ATLAS, can add constraints from physics & external data misaligment studies on pixels with HIP

20 Solving Large Degrees of Freedom ● Challenge: CMS and ATLAS have large systems to solve (100 k & 36 k Do. F) ● Formalisms require novel techniques ● Limiting factors: Size: Full ID needs O(10 GB) for handling the alignment matrices ● Precision: Matrices can have large condition numbers (compete with machine prec. ) ● Execution time: Single-CPU machines with non-optimized libraries take hours ATLAS: Currently solving using 64 -bit //-computing ⇒ full system was possible only last year! ● ● Solving full pixel (12. 5 k Do. F) on 16 nodes takes only 10 mins (7 hrs on Intel P 4, diagonalization) Work ongoing for improvements: already implemented MA 27 in Athena: takes 7 sec for 6180 Do. F, single-CPU CMS: ● Millepede-II using Min. Res was shown to solve 12 k Do. F in 30 sec in single-CPU! ● Generally, issues depend on the sparsity of matrices and other factors. Things get really complicated! MKU double precision (During datataking, a few mins performance differences in solvers may not be our bottleneck problem!) ~log 10 |AA-1 -I| ATLAS & CMS Alignment ● quadruple precision Inversion fails 0 -10 -20 0 20000 40000 60000 80000 N

21 “Weak” Distortional Modes. . MKU ATLAS & CMS Alignment Problem: Certain transformations leave 2 unchanged. Need extra handles to tackle these: • Requirement of a common vertex (VTX constraint), • Constraints on track parameters or vertex position (external tracking, calorimeters, resonant mass, . . . ) • External constraints (hardware systems, mechanical constraints, …). Easily incorporated in the formalisms (for ex, global 2) “clocking” R VTX constraint radial distortions (various) “telescope” z~R cosmics dependent sagitta X a b. R c. R 2 h dependent sagitta “Global twist” Rcot( ) Mass constraints, cosmics, E/p, charge dep

22 More on Weak Global Modes ATLAS & CMS Alignment Example “lowest modes” in PIX+SCT Global Freedom ignored MKU Ø Weak modes contribute to the lowest part of the eigenspectrum. ØThese deformations lead directly to biases on physics (systematic effects). Ø Such global effects already under study (lots of preliminary results, have no time to show all, so will sample in next pages!)

23 ATLAS & CMS Alignment MKU ATLAS “CSC” Challenge • Currently using “multimuons” data Level of applied misalignments: Modules = Level 3 with a realistic as-built geometry to Layers = Level 2 (barrel layers or disks) align the ID algorithms Subdetectors = Level 1 (whole barrel or EC) • Aim to test performance and Expected misalignments: • Modules: 30 -100 µm, 1 mrad understand needs for real data • Layers: 100 µm conditions • Silicon Barrels & EC: Up to few mm From detector assembling and installation: Misalignments largest on L 1 and smallest on L 3 ⇒ Alignment strategy: L 1 ⇒ L 2 ⇒ L 3 • Same misalignments are also used to check physics performances • Bs studies with misalignment: 14% less candidates reconstructed (B. Epp)

24 CSC: Algorithm Performances ATLAS & CMS Alignment • Check if the algorithms converge and improve residuals Global chi 2 Nominal 1 st Iteration 8 th Iteration Input Misalignment MKU • Check if efficiency and track parameters improve Robust Alignment Local chi 2 recover pixel

25 CSC: Welcome to the Real World MKU ATLAS & CMS Alignment • Improved residuals is only a part of the story. . • Are we able to see systematic effects (mostly weak modes) after alignment? Yes, as biases in track parameters • As the algorithms cannot fix these alone, use additional constraints • Transverse translations detected and already incorporated in algorithms: vertex/beam spot fit. Especially to be studied in pilot run. • Also apparent in p. T, mass and charge-dep. E/P handles

26 ATLAS & CMS Alignment ATLAS ID Alignment: CTB Performances y x z 6 PIXEL 8 (-1) SCT modules Modules (1 dead) • First real data from ID at H 8 beam in 2004 • Large statistics of e+/e- and (2 -180 Ge. V) (O(105) tracks/module/E), B-field on-off runs • Limited layout (6 Pi. X, 8 SCT, 6 TRT) • Results from various algorithms are being combined: reached a level sensitive to effects at a few microns! Overall residual resolution obtained: Pix residual sigma ~10 m, SCT ~ 20 m PIXEL MKU Before After Robust Excellent agreement

27 ATLAS SR 1 Cosmics: Performances • Before Alignment ATLAS & CMS Alignment • Surface (SR 1) runs in spring 2006: ~400 k Barrel cosmics recorded (22% of SCT, 13% of TRT) No B-field! No momentum! MSC important ~<10 Ge. V, need to deal with larger residuals than CTB Average Unbiased Residual Sigma [mm] MKU • Largest sample used ~200 k tracks Robust Global 2 Local 2 Helen Hayward Excellent assembly precision! Global 2 TRT+SCT: relative twist of SCT and TRT of 0. 2 mrad

28 ATLAS ID optical alignment (FSI) • ATLAS & CMS Alignment • • • Frequency Scanning Interferometry: Geodetic grid of 842 simultaneous length measurements (precision <1 m ) between nodes on SCT support structure. Grid shape changes determined to < 10 m in 3 D. Time + spatial frequency sensitivity of FSI complements track based alignment: – Track alignment average over ~24 hrs+. high spatial frequency eigenmodes, “long” timescales. – FSI timescale (~10 mins) low spatial frequency distortion eigenmodes -> weak global modes! Software principles already studied, implementation to be finalized! Barrel SCT Time months days FSI Tracks hours MKU minutes 80+(3 x[80+16])+(2 x 72)=512 End-cap SCT 165 x 2=330 seconds Spatial frequency eigenmode

29 ATLAS FSI on detector MKU ATLAS & CMS Alignment • ATLAS FSI barrel is mostly serviced and cabled in the pit (only waiting for endcap for final touch). • FSI will be used intensively before and during the early runs and the track-based alignment and FSI interplay will be tested. Stability of the detector will tell how frequent data needs taken during normal operation. Distance measurements between grid nodes precise to <1 m

30 ATLAS -spectrometer Alignment • Spectrometer: 1252 MDT chambers (708 Barrel, 544 Endcap) ATLAS & CMS Alignment Muon track will be measured with 3 drift tube chambers (~18 -20 layers) Requirement: 10% p. T resolution on 1 Te. V muon: sagitta of 500 μm measured with 50 μm accuracy Þmuon chambers must be aligned to 30 μm (Intrinsic resolution of a channel: 80 μm) MKU When Toroid is on, chambers will move by several mm => Optical alignment needed. hourly geometry changes expected. MDTs monitored for 9 chamber distortions, eg, elongation, sagging, . . 3 system with 3 point principle Florian Bauer, 4/9/2006, LHC Alignment Workshop BCam CCD Lens 53 mm 91 mm Spot target

31 ATLAS alignment Status MKU ATLAS & CMS Alignment • Optical alignment software validated at CTB, hardware installation underway. • 2 softwares: ASAP (barrel) and Ara. My. S (endcap) • Combine optical information with straight/High Pt tracks in global fit • Describe the 9 chamber deformations in the fit => 6 + 9 Do. Fs per chamber. • Handle up to 10 k Do. Fs both in the Barrel and Endcap • Run online with a latency of 24 h. => robust algorithms, automated dataflow, monitoring, use of Databases as IO • For alignment&calibration, a special L 2 trigger data stream is being setup • Misalignment studies show that the algorithms see the misalignments • Obtaining the required alignment is shown to take about 1/2 day, assuming parallel inner/outer chambers (ATLAS T&P week). Barrel End. Cap

32 ATLAS & CMS Alignment MKU CMS Hardware Alignment System • Components • Internal muon alignment • barrel (all chambers) • endcap (selected) • Internal tracker alignment (LAS) • TEC w. r. t. TOB; • TEC w. r. t. TIB. • Muon w. r. t. tracker (Link system) • Specifications • Tracker structures ~10 -100μm • Muon chambers at ~ 100μm • Muon vs tracker ~ 100μm • LAS: • monitor selected modules to get global alignment • 16+10+12 beams in total • Beams treated like tracks

33 ATLAS & CMS Alignment CMS Hardware Alignment System • CMS LAS has been used in parts successfully during reconstruction and is installed at the test centre at CERN for tests • Treats beams as tracks: nothing but another straight track fit! MKU • Full CMS hardware is 40 k parameters • Lots of software challenges, similar to track-based algorithms Transfer plate Clinometers Z-sensors Note: only small sample of analog sensors shown R-sensors DCOPS muon

34 Surveying the Detectors MKU ATLAS & CMS Alignment • ATLAS SCT Barrel photogrammetry survey was done in 2006 at SR 1 • measurements tell two faces appear to be rotated in opposite directions, hinting at twists of the complete barrel (order of 100 μm). • CMS results from including survey constraints in alignment shows improvement in residuals (similarly in ATLAS)

35 ATLAS & CMS Alignment ATLAS & CMS Tracker Status ATLAS: • Barrel tracker (except pixel) integrated in the pit and soon will take cosmics data. Pixels and endcap taking cosmics at surface. Endcaps installation end of may, pixels will go in mid-june. CMS: • Strip tracker complete and its 1/8 th is being read out. August onward, it will be completely in the pit and will take cosmics from mid-october. Installation plans for pixel is to be ready for data taking in 2008. ATLAS TRT+SCT Endcap CMS outerbarrel slice test (Feb 07) MKU 0. 1 mm

36 Conclusions • MKU ATLAS & CMS Alignment • • Both experiments have similar challenges and ideas for alignment, with different choice of optical alignment systems. Both experiments’ track-based alignment software in place, heavily tested, and providing proof of principles. They are at the cutting edge of today’s computing resources. We have been looking at real data already! Misaligned simulation studies underway. A lot has been learned, fixed, improved, but there is a lot more to do! We will be ready for collision data, however, full scalability needs to be proven with real collision data conditions: datastream from triggers, huge data samples, computing power, GRID-readiness, etc. The first collisions will be useful to exercise further the tools and understand the actual needs (time-scales, online monitoring, . . . ) rather than providing the final set of constants. Thankfully we do not need to reinvent most of the wheel, previous colliders suffered from similar symptoms. We need to be prepared for the unexpected, many issues upstream and downstream of alignment algorithms will exist and need to be understood: we cannot expect to obtain final module level alignment from day 1, but will likely nail down the global structures quickly. Please Stay Tuned!

37 ATLAS & CMS Alignment MKU Thanks • Ian Tomalin (CMS/RAL) for pointing me to the CMS information and answering my questions. • Pawel Bruckman (ATLAS/Oxford) • Jochen Schieck (ATLAS/MPI Munich) • Andrea Bocci (ATLAS/Duke) • Maria Costa (ATLAS/Valencia) • Numerous figures/slides borrowed from various talks, especially from the LHC alignment workshop of last fall. • Of course, thanks to all the alignment and detector teams of both experiments!

MKU ATLAS & CMS Alignment BACKUP 38

MKU ATLAS & CMS Alignment 39 ATLAS schedule

40 ATLAS & CMS Alignment ATLAS ID r view 107 cm 56 cm MKU 4 cm 30 cm

41 Robust Alignment: Concept MKU ATLAS & CMS Alignment Sum over neighbours, take correlations into account Sum over all modules in a ring Correct for change in radius

42 FSI + Track Alignment • How to include time dependency? 1. 2. MKU ATLAS & CMS Alignment 3. 4. FSI provides low spatial frequency module corrections at time ti , t 0

43 ATLAS & CMS Alignment MKU The challenge of putting it all together: Alignment data flow (Martin Weber)